Ariel Deschapell | Authors | Hydra Host

Ariel Deschapell

The Cloud’s TSMC Moment: Why The Largest Neoclouds Won't Own Hardware&nbsp; In 1987, the founding of Taiwan Semiconductor Manufacturing Company (TSMC) began a fundamental shift in the semiconductor industry by<a href="https://www.tsmc.com/english/aboutTSMC/dc_infographics_foundry#:~:text=computers%2C%20automotive%20electronics%2C%20the%20Internet,and%20have%20significantly%20improved%20our" rel="noopener noreferrer" target="_blank"> pioneering a “pure-play” foundry model</a>. Prior to this, industry heavyweights like Intel, Texas Instruments, and NEC Semiconductor were vertically integrated device manufacturers (IDMs), handling both integrated circuit (IC) design and manufacturing. TSMC changed that by specializing exclusively in manufacturing, enabling a new generation of fabless startups like NVIDIA to outsource this capability and specialize solely in IC design.&nbsp; The decoupling of the semiconductor business has had enormously disruptive consequences. Today, NVIDIA is the largest semiconductor firm by market capitalization and stands at 30x the capitalization of Intel. TSMC itself is 7x the size of Intel, and has become one of the most valuable and critical companies in the world, responsible for supplying virtually all of the cutting edge chips for giants such as NVIDIA and Apple.&nbsp; This specialization has facilitated enormous progress on both sides of the industry, which has directly led to the advancements which have facilitated the current AI boom. Now because of that same boom, similar economics are facilitating another decoupling and disruption further up the value chain.&nbsp;Today, cloud computing is experiencing the same critical threshold of complexity and increasingly divergent capital requirements that drove the decoupling of IC design and manufacturing in the late 80’s and 90’s. In the future most clouds, and eventually even the largest, won’t own any hardware at all.&nbsp; Understanding the Decoupling: Moore’s First and Second Laws&nbsp; Most are familiar with Moore’s Law: the number of transistors on integrated circuits doubles roughly every two years. However, Moore’s second law, less widely known but equally important, states that the cost of building state-of-the-art semiconductor foundries doubles approximately every four years.&nbsp;&nbsp;This exponential growth in capital requirements, and the parallel growth of design sophistication, naturally set the stage for decoupling and specialization. Companies found it increasingly difficult to master both ever more sophisticated chip design and the extraordinary complexity needed to manufacture them at scale cost effectively.&nbsp; Today, an almost identical dynamic is emerging in cloud computing. Cloud giants like AWS, Google Cloud, and Azure are vertically integrated, managing both massive server infrastructure and developing sophisticated software stacks on top of their own hardware. These big three are all doubling down on this strategy and attempting to take it even a step further by also designing their own in-house AI chips (<a href="https://www.wired.com/2016/05/google-tpu-custom-chips/#:~:text=This%20morning%20at%20Google%20I%2FO%2C,transforming%20the%20Google%20search%20engine" rel="noopener noreferrer" target="_blank">such as TPUs</a>) to further verticalize their stack.&nbsp; The first generation of AI-centric clouds (or “neoclouds”), like Lambda Labs, roughly follow this same vertically integrated playbook by building their own hardware capacity while pursuing their own abstracted software platforms in tandem. But because of increasingly similar economic pressures, this model stands to be disrupted, and efforts to vertically integrate even further stand to have massive blow back. The first to realize and execute on this effectively will be the ultimate winners just as TSMC and NVIDIA are today.&nbsp; The Emergence of “Pure-Play” AI Factories&nbsp; The accelerating sophistication and cost of GPU-powered data centers (referred to as<a href="https://blogs.nvidia.com/blog/dell-technologies-world-2024/#:~:text=Huang%20discussed%20the%20concept%20of,valuable%20data%20tokens%20distributed%20globally" rel="noopener noreferrer" target="_blank"> AI Factories by Jensen</a>) parallel the early history of semiconductor manufacturing facilities. Take the A100 GPU from NVIDIA which had an initial starting price of $20,000. Roughly 5 years later, the price of the soon to be released B200 GPU is close to $40,000. The power draw, meanwhile, has roughly quadrupled from 250W to 1000W, accelerating challenges in supporting infrastructure and operations. This rate of growing capex and complexity closely mirrors the dynamics of foundries in Moore’s second law.&nbsp;This isn’t the only pressure driving decoupling. As the cost and complexity of HPC infrastructure and AI Factories continue to escalate, so too does the pace of change in the world of software applications and tooling.&nbsp; As AI-driven software development accelerates, software iteration cycles will shorten dramatically, demanding frequent updates, adaptations, and specialized features to support its continued improvement and growth. If the rate of increasing costs and complexity of AI Factories seems to parallel that of foundries, then the speed of software iteration cycles seem poised to mirror the rapid progress of integrated circuit capacity in Moore’s first law.&nbsp; Though this is innately harder to measure, given recent advancements it doesn’t seem outrageous to suggest that software iteration speeds are on pace to roughly double every two years. Compare the state of the art in early 2025 to early 2023: Tools such as Cursor and Replit are now able to leverage current frontier models to reduce the barrier of entry for software development to a level never seen before. They enable the completion and deployment of working app POCs with a handful of prompts, while also turbocharging the abilities of experienced software developers. An<a href="https://github.blog/news-insights/research/the-economic-impact-of-the-ai-powered-developer-lifecycle-and-lessons-from-github-copilot/#:~:text=quantitative%20research%20study%20that%20developers,for%20productivity%E2%80%99s%20sake%20means%20nothing%E2%80%94we" rel="noopener noreferrer" target="_blank"> analysis by Github in May 2024</a> showed a 50% increase in developer productivity from using Copilot. These tools and capabilities are the worst they will ever be today.&nbsp; For the software side of the cloud business, this means demand for tooling and infrastructure to launch, manage, and produce this software will only continue to accelerate just as their own ability to iterate quickly on their own platforms accelerate as well. If they don’t, then their competition surely will instead.&nbsp; Just as AMD eventually<a href="https://arstechnica.com/gadgets/2008/10/and-then-there-was-one-amd-spins-off-foundry/" rel="noopener noreferrer" target="_blank"> spun out their foundry business</a> to go fabless despite fierce initial resistance ("<a href="https://clausaasholm.substack.com/p/real-men-have-fabs" rel="noopener noreferrer" target="_blank">real men have fabs</a>"), cloud service providers will begin to realize that AI hardware investment and cloud software development require fundamentally different expertise, investment profiles, and intense areas of focus.&nbsp; Unlocking a New Generation of Neoclouds&nbsp; Future neoclouds will be forced to specialize exclusively in software and user experience to beat and stay ahead of the competition, by offloading hardware investment and management to dedicated AI Factory providers. These AI Factories will function similarly to TSMC, by taking on the burdens of intense capex and reducing costs by sheer economies of scale. They will provide access to state of the art infrastructure through flexible bare metal APIs, freeing neoclouds to invest their resources entirely into differentiated user experiences and software innovation at higher levels of abstraction up the cloud stack.&nbsp; Such AI Factory businesses will face early challenges similar to those TSMC encountered, which lacked large obvious early customers as the major semiconductor firms already possessed in house foundries. But this didn’t particularly concern founder and CEO Morris Chang. Chang had previously served as an executive at Texas Instruments, where he worked with numerous engineers who dreamed of one day launching their own chip companies. Raising the capital to do this was extremely prohibitive however, with the cost of building an in house foundry quickly escalating into the hundreds of millions.&nbsp; By creating an independent foundry that they could outsource to, Chang reduced the barrier of entry substantially, enabling a generation of fabless chip startups to enter the market without prohibitive infrastructure investment needs. In a rare interview on the Acquired podcast, Morris Chang recollected the first time a then nearly bankrupt NVIDIA had reached out to him in the 90s: “I had always told our sales people we should never be negligent in talking to future customers, even if the customer seemed to be a very small one”.&nbsp; This approach paid off. NVIDIA, Qualcomm, Broadcom, Mediatek, P.A. Semi (acquired by Apple to form their in house chip division) and countless others flourished and reshaped the industry precisely because they were able to outsource to TSMC and other foundries adopting the pure-play foundry model.&nbsp; Reducing Barriers to Entry&nbsp; The same economic headwinds are now pushing upstream full force for cloud platforms. The capital and expertise required to operate GPU clusters and AI Factories are quickly becoming prohibitive. In parallel, the demands for the software side of the cloud business are only beginning their dramatic acceleration as well.&nbsp; Neoclouds will have to keep up with the demand for ever better and more expansive tooling as the barriers for producing software continue to decrease. Those that are able to focus purely on software and customer-facing innovation are poised to be far more successful, given they no longer need to shell out capital, time, and focus on an underlying hardware business.&nbsp; The emergence of dedicated AI Factories that can enable this will dramatically lower barriers to entry just as TSMC did, spawning new specialized cloud companies, each innovating faster and more creatively than today’s vertically integrated providers. Among them, several companies will inevitably scale into formidable players, reshaping the competitive landscape and potentially disrupting established hyperscalers just as Intel is nearly an afterthought today.&nbsp; TSMC's choice to specialize transformed the semiconductor industry and, with it, the world. The cloud industry’s TSMC moment is now happening as we speak, as a direct consequence of the previous specialization in the semiconductor industry. Vast AI, Runpod, Salad, and SF Compute are just a handful of startups attempting to build software only cloud businesses on top of bare metal providers, while my company Hydra Host is taking the approach of enabling independent data-centers to transform into specialized AI Factories with our turnkey bare metal solution and bundled design, procurement, and financing services.&nbsp; The decoupling is underway. The only questions are how quickly will it play out, and will the hyperscalers be able to adapt? Or will they become the new generation of IDMs like Intel and NEC Semiconductor before them?

TSMC's choice to specialize transformed the semiconductor industry and, consequently, the world. The cloud industry's TSMC moment is occurring now, driven by semiconductor specialization. Companies like Vast AI, Runpod, Salad, SF Compute, and Hydra Host, which transforms independent data centers into specialized AI Factories, exemplify this shift. 

The Cloud’s TSMC Moment: Why Neoclouds Won't Own Hardware

Recent chip export restrictions designed to deny advanced hardware to the Chinese Communist Party have swept up several other countries in the process. Most notable are the Gulf States of the Middle East- such as the UAE, Saudi Arabia, and Kuwait.With the rapid rise in popularity of Deepseek (almost certainly trained on restricted H100s), these export rules are facing renewed scrutiny. Enforcing direct prohibitions on sales to China is simple, but preventing subsequent resale through other jurisdictions is far more complex. To address this, the heavy-handed approach used by the previous Biden administration severely limited exports to many partner and neutral countries- a policy that remains leaky and ultimately harms U.S. interests. To understand why requires asking a simple question. Do we want the rest of the world to buy American or Chinese? Our restrictive export measures effectively subsidize the Chinese chip industry by knocking American competition out of key markets and pushing neutral or friendly nations deeper into China’s technological and economic orbit. It’s a case of cutting off our nose to spite our face.Fortunately, the choice is not between accepting the current policy or abandoning all restrictions on critical hardware altogether. We can apply better technological solutions to enable effective export controls without inflicting so much collateral damage—and we must do so swiftly. <h2>The Gulf States: Key to the Compute Race</h2> No place illustrates this policy dilemma more clearly than the Arabian Gulf. Historically renowned for its strategic value- from ancient trade routes to its modern status as the nucleus of global oil production- the Gulf region is on the verge of becoming a central player in AI infrastructure. Countries like Saudi Arabia, the UAE, Kuwait, Oman, and Qatar have invested heavily in power generation and internet infrastructure. As a result, they not only surpass the baseline for AI development but could become the most consequential partners worldwide in the emerging “compute race.” Like the transition from early gasoline engines to mass-market automobiles where every technological leap fueled even higher demand for oil- each new AI breakthrough (à la Deepseek) will require exponentially more compute power and thus more energy to run advanced workloads at scale. All this means that raw energy output is quickly becoming the limiting factor for developing high-end compute. The United States has already tapped most “low-hanging fruit” sites for data centers, forcing some hyperscalers like AWS and Azure to even invest in <a href="https://www.datacenterdynamics.com/en/opinions/hyperscaler-nuclear-power-ambitions-heat-up/" rel="noopener noreferrer" target="_blank">restarting nuclear plants.</a> In this context, the Gulf States’ energy capacity stands out dramatically. As of 2023, Saudi Arabia, the UAE, and Kuwait collectively produce around 16 million barrels of oil per day- nearly 17% of global output. Their proven reserves are even higher with plenty of headroom to ramp up production. In addition to fossil fuels, these nations are pouring resources into renewable energy projects such as solar further expanding their net energy surplus. As a whole, the Gulf region is the largest net exporter of energy in the world, making them ideal partners for large-scale AI endeavors that demand vast, stable power supplies.In the near term, it will become increasingly apparent the Gulf region will rapidly become the most consequential provider of AI infrastructure capacity just as it is the most important producer of oil today. The only question is with whom will it partner with to develop this capacity: American businesses or the rapidly growing Chinese tech sector. Currently, the U.S. leads in advanced chip capabilities, but if we continue to withhold those capabilities from a region with such vast potential, China will not hesitate to seize the initiative to boost its own capabilities. <h2>A Technology Problem</h2> So how can we enable key international partners effective access to American chips and hardware, while ensuring they do not ultimately make their way to China once delivered? Simple. By bundling deployments of restricted hardware with tailored DC management software like Hydra Host's Brokkr platform, we can ensure that delivered GPUs are actually installed and remain on site in the intended destination country. The Brokkr platform provides a live source of truth for all these deployments to OEMs and other stakeholders, providing ongoing and automated compliance monitoring. If GPUs are disconnected from the authorized data center site in which the software is installed, it triggers an immediate alert providing the assurance necessary to relax export concerns. By implementing smarter policies and leveraging targeted technology to allow Gulf States to purchase American hardware without the risk of resale leaks, we can strengthen U.S. influence, bolster domestic chipmakers, and maintain our technological edge—all while countering China’s growing presence in a region critical to the future of AI. This approach doesn’t mean abandoning security or ignoring export controls; it means using more nuanced tools to deny our most advanced technology to adversaries without undermining partnerships and ceding these markets to China.

The Middle East is set to become a global AI hub, but US chip restrictions could push it toward China. Learn how smarter policies & Brokkr tech can secure US interests.

Advancing American Interests with GPU Exports to the Middle East

In the 1950s, McDonald’s revolutionized the fast-food industry, expanding from a single restaurant to over 200 locations in just five years. This kind of growth was unprecedented, and the secret lay in Ray Kroc’s now ubiquitous innovation: franchising. Kroc created a system to empower capable entrepreneurs with their own capital to scale what worked and piggyback on the parent company's brand credibility, expertise, resources, and distribution. The franchise model enabled rapid expansion while maintaining standardized operations, procurement, and delivery. Across industries ranging from dining to automotive care to shipping and logistics, the franchise model has since devoured many smaller competitors who simply lacked the scale and flexibility to compete. Today, a similar transformation is playing out in real time for AI factories—high-performance computing (HPC) data centers optimized for artificial intelligence workloads. These GPU-focused facilities, the backbone of AI development, face challenges that echo those of early McDonald’s: high capital needs, ever-changing operational complexity, and scalability bottlenecks. Here's how franchising can offer the same breakthrough for AI factories as it has in many other industries. <h3>The Challenge of Monetizing AI Factories</h3>Building an AI factory is no small feat. These data centers require significant capital outlay for real estate, infrastructure, and high-performance servers. And that's only the beginning. Once the facility is operational, the real challenge emerges: distribution and monetization.For most data center operators, setting up hardware is the easy part. Creating and deploying an entire cloud platform so customers can actually utilize that hardware, and acquiring those customers, present an entirely new set of challenges that's wholly alien to the traditional data center industry. As a result, even well-funded data centers often struggle to attract sustained demand or stand out in an increasingly crowded market. A recent example highlights this reality: data center Equinix acquired Packet, a promising bare metal cloud company, for $335 million in 2020. Despite high expectations, Equinix recently announced the sun-setting of the renamed Equinix Metal Cloud offering to refocus on its core colocation business.The Packet acquisition and subsequent scale up investment by Equinix failed, resulting in a massive write-off. The example of Equinix Metal underscores the complexities of running a cloud business alongside a traditional data center business. Each has vastly different operational and customer needs, and even a company with Equinix's vast resources couldn't effectively bridge the gap.This presents a critical issue. The global AI ecosystem benefits from decentralization, where a diverse range of independent data centers can ensure competition, innovation, and resilience. However, without a scalable and replicable growth model, the industry risks GPU power becoming concentrated in the hands of a few dominant players, stifling competition and creating chokepoints for innovation. That is of course, much like the traditional cloud. <h3>Franchising as a Solution</h3>Franchising offers entrepreneurs the opportunity to be at an industry's cutting edge day one. Just as McDonald’s created a template for entrepreneurs to replicate its operations, AI factory franchising will provide a standardized blueprint for data center expansion and operations using near-turnkey software and ongoing full stack support. Through a franchise model, individual operators and investors can deploy their expertise and resources in developing their own AI Factory projects, while benefiting from shared procurement operations, financing relationships, software infrastructure, distribution channels, and of course an existing customer base. Contrast this with today's non-standardized approach, wherein aspiring data center operators must reinvent the wheel and spend precious time and resources rediscovering existing best practices to finance, procure, and monetize their chips. Here's how Hydra Host is pioneering data center franchising:<ol><li>Faster, Cheaper Infra Procurement: Small, independent data centers often struggle to acquire the latest GPU models as quickly as their large cloud company competitors. When they do get a quote, they often lack the leverage needed to access fair pricing. Hydra takes care of pricing leverage, providing small data centers the benefits of scale instantly. Through existing relationships with companies like NVIDIA and various infrastructure and OEM partners, Hydra Host aggregate orders and passes those benefits directly to our DC partners.</li><li>Easy-to-Use, Standardized Infrastructure: Hydra Host's Brokkr platform features a bare metal management and provisioning system, providing a source of truth for server management and status, as well as a way to monetize bare metal servers in a secure and automated way. Bare metal skips the need for a hypervisor or containerized abstraction layer, reducing complexity and increasing stability and performance, making it a premium choice for AI model and platform companies. It also means servers on Brokkr are compatible and can be pushed to third party marketplaces, such as Runpod and <a href="http://salad.com/" rel="noopener noreferrer" target="_blank">salad.com</a>, to supplement Hydra driven demand when needed.</li><li>Centralized Distribution: Creating a competitive platform is one of the biggest hurdles independent data centers face. Brokkr takes care of this and acts as a unified user interface and API for all of our DC partners, delivering end customers a reliable, centralized source of supply with a predictable and standardized purchasing experience.</li></ol> Hydra Host solves provisioning for its clients, handling utilization and passing along the benefits of a its standard and ecosystem so data centers only have to concern themselves with on-premise day to day operations and maintenance. <ol><li>Outsourced Customer Acquisition and Sales: Just as McDonald’s centralized its branding and marketing, Hydra is constantly sourcing paying customers and funneling clients to data centers integrated with Brokkr. Individually, independent operators can't compete with industry giants long term. But as part of the Brokkr umbrella and ecosystem, data centers can combine to quickly scale to more locations and provide more diverse offerings and configurations than any centralized cloud provider can hope to match.</li><li>Financing: Hydra Host exploits its visibility into compute and hardware pricing, as well as its procurement channels, to provide its clients advantageous and rapid access to equipment financing. Hydra's finance partners can underwrite with greater precision, confidence, and speed than any other financiers in the space. This means Hydra's data center clients enjoy better and faster financing than they could access anywhere else.</li></ol> <h3>Why It Matters</h3>Franchising AI factories wouldn’t just enable faster growth; it could democratize the entire AI infrastructure market. A distributed network of AI factories, each independently owned but operating under a standardized system, could help ensure that GPU power isn’t concentrated in a handful of companies or geographies, and enable much more experimentation and innovation. Franchising allows investors and operators to scale what works instead of reinventing the wheel. Data centers are overdue for a turnkey, just-add-capital approach allowing them to focus on scaling instead of fighting for relationships and rediscovering what their competitors already discovered years before. <h3>The Road Ahead</h3>Franchising isn’t a silver bullet, and there are significant challenges to overcome. The balance between standardization and local flexibility can be tricky to manage. However, the potential benefits are significant. Franchising has historically demonstrated its power to improve the odds of success in competitive industries. For example, while only ~20% of independent restaurants survive beyond five years, top franchises like McDonald’s boast survival rates of over 90% in the same timeframe. AI factories will follow a similar dynamic. Standalone data centers often face steep challenges, including building platforms, acquiring customers, and navigating competitive pressures. A franchising model for AI factories delivers the same advantages franchises provide in other industries: a proven playbook, operational support, and the strength of a unified brand. By empowering local entrepreneurs with these tools, franchised AI factories unlock a path to greater success, ensuring more resilient and distributed AI infrastructure worldwide.

Franchising AI factories wouldn’t just enable faster growth; it could democratize the entire AI infrastructure market. A distributed network of AI factories, each independently owned but operating under a standardized system, could help ensure that GPU power isn’t concentrated in a handful of companies or geographies, and enable much more experimentation and innovation. 

AI Factories In A Box: Why Franchising is the Future of Data Centers

<a href="https://bitcoinmagazine.com/technical/compute-maximalism-the-symbiosis-between-bitcoin-mining-and-ai" rel="noopener noreferrer" target="_blank">Initially published at Bitcoin Magazine</a> The story of human progress can be simplified into the story of increasing energy utilization. We harness energy to create order, both in terms of biology and society. Energy surpluses allow for every form of wealth creation, which in turn produce new technologies to effectively harness yet more energy. This truth has inspired famous concepts such as the Kardashev scale, which measures civilizations by their ability to harness energy resources towards useful ends. Compute is a natural continuation of this endeavor. Modern digital technology transforms ever greater quantities of electricity into advanced value-creating processes. The most recent surge in demand for compute has come from two technologies in particular: bitcoin mining and, more recently, high performance compute (“HPC,”) in particular the use of Graphics Processing Units (“GPUs”) for artificial intelligence. The meteoric rise in energy consumption by these technologies have raised many questions: What impact will these power-hungry technologies have on our energy systems? Given their mutually voracious use of energy, what interplay will they have with each other? What do these developments mean for humanity? We explore the essential characteristics of these respective technologies, and how they offer alternative markets for excess power that can in fact improve the efficiency of energy systems. Based on this exploration, we also argue that bitcoin mining and HPC are complementary rather than competitive. As we’ll see, their respective trade offs offer a symbiotic ability to maximize the value created from energy resources, which in turn benefits society as a whole. In short, we argue for compute maximalism. <h2>Energy</h2> Modern technology depends on energy converted from a wide array of sources into electricity, and this comes with certain challenges and tradeoffs. The primary of which is limited portability. This is due to several simple realities. Electricity requires a grid, essentially a massive series of circuits which transports energy in real time. The grid must remain in balance, meaning the generation must roughly equal demand at any point in time. This is difficult for two reasons: First, energy resources are not always conveniently distributed, have long lead times to develop, and have varying degrees of dispatchability. Second, transmission as well as storage are both expensive, have similarly long lead times, and experience inherent inefficiencies. An estimated <a href="https://demo.openeclass.org/modules/document/file.php/DEMO-A1153/%CE%A3%CE%B7%CE%BC%CE%B5%CE%B9%CF%8E%CF%83%CE%B5%CE%B9%CF%82%20%CF%80%CF%81%CE%BF%CF%82%20%CE%BC%CE%B5%CE%BB%CE%AD%CF%84%CE%B7/EFFICIENT%20ELECTRICAL%20ENERGY%20TRANSMISSION%20AND%20DISTRIBUTION%20.pdf" rel="noopener noreferrer" target="_blank">8-15%</a> of electricity is lost to transmission and distribution losses by the time it reaches local consumers, and this figure is even higher for long term battery storage. The result is that it will always be cheaper and more efficient to consume generated electricity immediately at the source, than to transport it over time or space. As such, the most efficient solution isn’t to more widely and inefficiently transport electricity to where it can be used, but to move use cases to the electricity. Compute is an ideal use case for such excess electricity because it is power dense, largely portable, and scalable; we are yet to find a limit on our demand for compute.Meanwhile “meatspace” constraints are strong limiting factors for legacy forms of energy sinks like aluminum smelting and manufacturing. Bitcoin mining in particular has emerged as such an ideal use case for local surplus power, providing a dispatchable and revenue-generating load to balance the grid. More recently the demand for High Performance Compute, in particular GPUs, is also having unignorable impacts on the utilization of energy as well. Many are expecting these two technologies to compete over the same energy resources, but as we explore the characteristics of each, the potential symbiosis will become self-evident. <h2>Bitcoin Mining</h2> Bitcoin mining can be thought of as a permissionless energy sink. Bitcoin’s proof-of-work consensus mechanism amounts to proof of energy-intensive computation. Miners must perform this energy-intensive computation to create new blocks of transactions, thereby earning bitcoin as a reward. It is this proof-of-work that provides global settlement assurances in a decentralized and permissionless way. In practice, this looks like millions of computers (these days, application specific integrated circuits or “ASICs”) running in bare bones data centers around the world. One of the beautiful things about bitcoin mining is its permissionless nature; anyone anywhere in the world can plug in an ASIC. In effect, Bitcoin allows miners around the world to participate in a global energy market; whoever has the lowest cost of power has the highest margin. This global decentralized network is part of the reason why Bitcoin’s adoption has steadily continued, as people seek a new monetary and financial system that is active 24/7, lacks a single point of failure, and sidesteps the perverse incentives of politically captured central bank monopolies. Bitcoin mining is distinguished by the following characteristics relative to GPU/HPC infrastructure: <ul><li>No customers</li><li>No customer acquisition</li><li>No support</li><li>High interruptibility</li><li>Low operational complexity</li><li>Low connectivity requirements (Less than 100MB/s)</li><li>Low margin (generally)</li></ul> <h2>HPC</h2> Data center GPUs are the latest form of HPC, the demand for which has exploded in the previous 2 years due to quickly escalating interest in AI/ML breakthroughs which rely on it. These technologies have unlocked whole new categories of digital operations and functions which were not previously possible, with the resulting use cases only just beginning to be explored. The sudden explosion in interest in these technologies has quickly made NVIDIA, the leading manufacturer of the underlying GPUs, the most valuable company in the world. Initially this sudden spike in demand created an intense bottleneck in the production of sufficient units of GPU itself. This however was temporary and over time continues to be alleviated by increased production, with focus quickly switching to a new bottleneck: Data center rack space with cheap power. The result has been an explosion in new data center build outs, wherever a large amount of steady power can be sourced. This has brought GPU infrastructure into competition with Bitcoin mining in many areas with excess local power. Relative to Bitcoin mining, GPU/HPC has these is distinguished by these characteristics: <ul><li>Customers</li><li>Customer acquisition</li><li>Customer support</li><li>Low interruptibility</li><li>High operational complexity</li><li>High connectivity requirements (10 - 100GBs)</li><li>High margin (generally)</li></ul> <h2>Complimentary Competition</h2> The demand for both Bitcoin and AI/ML technologies has taken off in the last decade, a testament to their utility to society. This demand has led to the proliferation of their respective compute resources. To reduce operating costs, both markets seek excess power to utilize as it tends to be cheaper. This naturally resolves some of the grid inefficiencies discussed above, but it does mean that data center builders and operators will find themselves asking which form of compute to support and invest in for the same amount of available power. Both forms of compute are energy intensive and relatively location-agnostic (barring legal or jurisdictional considerations beyond the scope of this paper) bringing them into seeming competition, but they can in fact be highly complementary tools for maximizing utilization and profit from such excess or stranded electricity. GPU workloads have higher operational complexity, and low interruptibility, as well as higher upfront capital investment. That makes it a poor choice for taking advantage of transient surpluses of power, such as the peak window of energy production by solar panels for instance. Unlike bitcoin mining, GPUs have customers, who are typically sensitive to issues such as uptime and availability. There are exceptions, such as spot instances and frameworks which can failover from such instances, but generally speaking due to the existence of a customer the interruptibility tolerance of GPU infrastructure will never match that of bitcoin mining. Coupled with the higher capital costs and complexity, in these situations we can expect bitcoin mining to continue to grow and dominate as a highly flexible, dispatchable load to the grid. Consistent excesses in power on the other hand, such as a largely fixed delta between the base generation capacity of hydropower or nuclear sites and their surrounding consumption, are ideal opportunities for GPU infrastructure to close the gap and establish new baseline consumption and equilibrium. These situations favor the low interruptibility of GPU infrastructure, and justify the added expenditure and operational complexity in order to secure substantially higher revenues. So long as the supporting bandwidth is available to facilitate GPU workloads (at least 10GB/s, ideally 100GB/s), these sites will always provide more profit opportunity than if allocated exclusively for bitcoin mining. <h2>Hybrid data center Strategies</h2> There are also strategies which can utilize both technologies in tandem to maximize revenue and return on investment. First, bitcoin mining could be used as an initial load for energy resources before the site is suitable for high performance compute. Examples include: (1) using semi-portable modular bitcoin mining data centers to monetize power while the remaining infrastructure for an HPC data center (redundant power/internet lines, buildings, backup energy systems, etc.) is built; or (2) pioneering stranded energy resources with bitcoin mining, some of which may eventually be used for HPC. In fact, <a href="https://bitcoinmagazine.com/tags/core-scientific" rel="noopener noreferrer" target="_blank">Core Scientific</a>’s recently announced deal with CoreWeave could be viewed as an example of this occurring in the wild, as bitcoin mining led to the development of a large substation and data center shell that would eventually be used for HPC. A second, more advanced strategy is to co-mingle HPC and Bitcoin mining workloads in tandem, using Bitcoin mining as a counter weight to balance fluctuations in HPC workload power draws. While HPC loads require reliable power, “inferencing workloads” which host production AI/ML models can fluctuate based on levels of real time use by users, leading to typical cycles of high activity and power consumption and low activity and lower power consumption. To date, the value for such HPC has substantially outweighed any inefficiencies from fluctuating power use, but the highly flexible and interruptible nature of Bitcoin mining can be used to provide stable power draw and in turn lower effective power rates, in addition to providing additional revenue for the data center overall. Some are describing this strategy as a “mullet data center,” with AI in the front and bitcoin in the back. While it is still early, this approach promises to utilize the best of both HPC and bitcoin mining to offer the most value maximized data center deployments possible with current technology. <h2>Industry Implications</h2> Until recently, the data center industry has been dominated by colocation providers. These providers build the facilities used to host industrial servers, and lease out space, power, connectivity, and sometimes the servers themselves to tenants. Traditionally, the majority of these tenants have been large enterprise and hyperscale cloud providers. In many cases these hyperscale and enterprise tenants have also built their own data centers to support their own growth. Since roughly 2017 bitcoin mining has truly entered the picture at an industrial level, with entire data center complexes being built solely to support Bitcoin mining in areas with extreme deltas in produced and consumed electricity. Now in 2023 and 2024 we’ve seen shifts in the market even more notable and disruptive. With the explosion in demand for GPU infrastructure, many former colocation focused data centers have ventured into buying and hosting this GPU infrastructure themselves. Meanwhile hyperscalers are moving behind-the-meter to co-locate with large baseload power plants, seeking cheap reliable power for the new surge in HPC demand. This is particularly notable as intermittent renewables have been the most popular form of generation in recent years, primarily due to government subsidies. We anticipate the following: 1. Continued increase in energy demand for both forms of compute. 2. New data center construction as the next bottleneck of expanding HPC footprints, with large swaths of bitcoin mining facilities being repurposed for higher margin use cases. 3. Mining hardware will relocate to the fringes, seeking remote locations and variable inefficiencies that HPC workloads are ill-suited to monetize. 4. Co-mingling of both bitcoin mining and HPC in “mullet data centers” will leverage the high revenue potential of HPC and the flexible nature of bitcoin mining to effectively balance power draw and local grids, while outcompeting traditional data center strategies. <h2>Conclusion</h2> When new power-hungry technologies emerge, there is often concern about their energy utilization and its externalities. Bitcoin mining and HPC are no exceptions with politicians and arm-chair technologists alike crying for their mitigation or control. But such voracious technologies represent the natural trend of human progress. In addition to the self-evident utility provided by the Bitcoin settlement network and AI/ML workloads, we can demonstrate that they can be deployed in ways that efficiently maximize new and existing energy resources to useful economic ends.

Drew Armstrong

Discover how Bitcoin mining and GPUs drive energy efficiency and progress. Learn about hybrid data centers and strategies for compute maximalism.

Compute Maximalism: The Synergy of Bitcoin Mining and AI | Hydra Host

<h2>Brokkr | Hydra Host's High-Performance Computing Marketplace</h2>

Today the Hydra Host team is excited to announce Brokkr, a two sided marketplace for high-performance computing. The Brokkr platform aims to connect users with an extensive network of computing resources, offering unparalleled optionality and flexibility when it comes to compute heavy workloads. From scientific research and data analytics to artificial intelligence development and simulations, Brokkr matches users with the power they need to tackle complex tasks efficiently and effectively.&nbsp; "Brokkr" is a reference not only to the platform's function as a compute brokerage or marketplace but also to the mythological Norse figure, known as a skilled blacksmith and metalworker. This intentional terminology emphasizes Hydra Host's focus in designing and building Brokkr: enabling the provisioning of bare metal servers from any data center, anywhere.&nbsp; By addressing bare metal provisioning, Brokkr paves the way for aggregating compute resources from providers worldwide, irrespective of size, location, or existing provisioning or API availability. For end users, this translates to a gateway and the first genuine marketplace for compute. For data centers, it enables them to monetize and optimize infrastructure utilization by joining the Brokkr network with no need for provisioning expertise or marketing prowess.

<h2>Understanding the Benefits of High-Performance Computing with Brokkr</h2>

High-performance computing is swiftly emerging as a vital instrument across industries, from scientific research and engineering to financial modeling and artificial intelligence. Yet, these computing resources are capital-intensive, and sourcing, acquiring, and maintaining them can be an intimidating undertaking for many organizations. In practice, this often means navigating disparate supply chains and providers or becoming excessively dependent on a single provider. Brokkr introduces a novel approach by furnishing a centralized marketplace where users can effortlessly access and provision high-performance computing resources. Brokkr allows you to harness a growing network of providers and data centers to find the best match for your specific computing needs. Regardless of the provider you choose, there's no need to concern yourself with managing multiple accounts, billing, or learning and integrating new APIs. Hydra Host standardizes all these aspects to minimize operational overhead and ensure a smooth and vendor-neutral experience for users.

Hosting has been largely monopolized by a few major hyperscaler providers for over a decade. However, in the past few years, the landscape has been changing. A rapidly evolving digital environment has amplified the need for dramatically more compute, and compute in more places than ever before. From the unprecedented demand for cost-effective GPUs for AI/ML workloads to an escalating focus on edge computing and optimized latency, the hyperscalers are struggling to keep pace. This situation stems from the inherent unsustainability of the hyperscaler model, which has been living on borrowed time. By creating isolated ecosystems and maximizing vendor lock-in, they've endeavored to insulate themselves from competition and camouflage this simple reality: compute is a commodity, and ought to be treated as such. The Brokkr platform seizes this evolving shift in the industry by offering a distributed alternative to the hyperscaler model, thereby enabling independent providers with unique insights and localized advantages to bring them to the table. With Brokkr, users seeking the highest performance, lowest cost, or least latency compute for their respective use cases will be able to discover and provision it seamlessly, from anyone and anywhere.

The key to Brokkr lies in our growing network of data center partners. While there are a plethora of multi-cloud platforms and hosting solutions available, none of them permit independent data centers to easily offer and monetize their resources. Brokkr bridges this gap by incorporating bare metal provisioning capabilities, an essential step towards commoditizing compute. From the outset, our focus has been to support independent data centers, enabling them to concentrate on their core competency, i.e., setting up and maintaining physical hosting infrastructure, by resolving provisioning and customer acquisition challenges. Through our dialogues and industry research, we've interacted with several independent data centers and gained a first-hand glimpse at emerging industry trends poised for acceleration. This includes the modular DCs established and managed by one of our partners, EnergyX. Their approach allows for industry-leading construction and setup times, and the flexibility to deploy to any location with superior economic factors, such as latency, power cost, or environmental cooling. Brokkr aspires to make innovative approaches like EnergyX's accessible and transfer the advantages they offer over traditional hyperscaler resources directly to end users. If you're a data center or a supplier of computing, we welcome you to discuss listing your resources on Brokkr.

Brokkr is still in its initial stages, and there's lots of work to be done. We're looking for early testers to help us iterate quickly and effectively. Below is a list of features currently on our roadmap, which you can expect in the comings weeks and months: <ul><li>A growing list of inventory from our hardware providers</li><li>Fine-grained filtering and search capabilities to allow users to easily find the specific compute resources they need based on performance, cost, and latency requirements.</li><li>Supply chain visibility: It's no surprise GPUs are in high demand, with most high-end models like the new H100 being sold out before they hit the import docks. Currently, buyers of high-end computing resources face challenges in tracking the supply chain and availability of these resources, and often have to do so by juggling several email threads and contacting multiple vendors. We aim to address this problem in Brokkr by enabling suppliers to list upcoming inventory available for pre-order and enabling interested buyers to browse this upcoming inventory all on one centralized dashboard for shopping and pre-ordering.</li><li>Terraform provider: In addition to our web interface and API, we are developing a Terraform provider for Brokkr. This will allow users to easily provision and manage their compute resources using modern infrastructure as code patterns.</li><li>Advanced billing dashboard: We understand that managing and tracking billing can be a complex task, especially in the world of high-performance computing. But it doesn't have to be. With Brokkr we are working to provide an easy-to-understand high-level billing experience, with the ability to zoom in to any desired level of granularity for accurate tracking and analysis. Compare and project compute resource and pricing options, and gain insights into resource utilization and costs, all in one comprehensive dashboard.</li><li>Secure boot and other security add ons: To ensure the highest level of security for our users, Brokkr will provide secure boot capabilities and other features to ensure the hardening of critical workloads regardless of where they are deployed.</li><li>Spot market: Utilization is one of the biggest problems when it comes to high performance compute. High end hardware is a massive investment, and making sure the most is being made out of it is critical. That's why Hydra Host is planning on extending Brokkr with a spot market for interruptible workloads, tailored to high end GPUs, in 2024.&nbsp;</li></ul>

<h2>Getting Started: How to Join the Brokkr Community</h2>

For both data centers and end users, all Brokkr usage requires manual on-boarding with the aim of working closely with users to identify potential bugs and gather feedback for further improvement. Whether you're a data center or end users, you can get started by registering below.

The Hydra Host team is looking forward to rapidly improve on the Brokkr platform, and add more compute partners and inventory. Be among the first with us on this journey, and enjoy limited time access to extremely cost effective Hydra managed CPU resources, and Founder level engagement and support.&nbsp;

Discover Brokkr by Hydra Host—a marketplace connecting users with high-performance computing resources for AI, data analytics, and more.

Introducing Brokkr

It is difficult to overemphasize how much “The Cloud” has become a mainstay of the Silicon Valley zeitgeist since its meteoric ascendancy from 2008 onward. Consequently, it may seem like an impossibly far-fetched idea that the growth and evolution of the cloud was a phenomenon as artificial as the interest rates which fueled it. The public cloud has witnessed roughly 2350% growth between the years 2008 and 2020, with even more headwind made immediately following the COVID-19 response and global lock downs. This particular spike in growth has been reasonably traced to the skyrocketing demand for digital services from an unprecedentedly shut-in population. But here’s something else this period had in common with the initial 2008 takeoff: historically low interest rates. Federal Funds&nbsp;Rates Since the 2008 financial crisis, central banks around the world have maintained historically low interest rates to stimulate economic growth and recovery. These artificially low interest rates have had an expansive range of effects on the economy, including influencing the growth and evolution of the public cloud. Recently, interest rates have experienced a sudden and sharp reversal. Correspondingly, the decade-long trend of ceaseless migration of digital infrastructure to hyperscalers has also witnessed its first major slowdown and vocal push backs. This article will explore the causative connection between low interest rates and the expansion of public cloud services. Furthermore, we’ll be discussing the alternatives to the public cloud paradigm that are poised to surge in growth, and what that means for various players in the industry. Interest Rates and the Public Cloud Growth The tech industry does not operate in a vacuum and is heavily influenced by macroeconomic forces in simultaneously subtle and dramatically impactful ways. We’ve already pointed at the correlation between the rapid growth of the cloud and historically low interest rates beginning in 2008. But to explore the causative mechanisms at play between these concepts, we need to ensure we understand the economic backdrop of the cloud’s development.First and foremost, interest rates represent the cost of borrowing money, which is ultimately determined by the Federal Reserve and other central banks worldwide. When interest rates are pushed artificially low, liquidity is abundant and inexpensive, resulting in both profound and non-obvious downstream impacts on the behaviors and trends of entire economic sectors. In such conditions, companies and investors are strongly incentivized to pursue riskier investments and are more likely to take on debt and fundraising to finance their ventures. The market begins optimizing for growth, often and even inevitably at the cost of capital efficiency, fiscal discipline, and long-term strategic planning. This is due to the reality that if a company is not leveraging the lower interest rates and increased credit availability to focus on growth while their competitors are, they are unlikely to survive. Consequently, all market participants eventually adopt similar optimization strategies. This leads to the development of entire industries on shaky foundations that would have been impractical and imprudent outside of the influence of extraordinary macroeconomic conditions. It’s only once conditions return to the norm, that the decision making that seemed sound and rational previously is met with a harsh dose of reality. When interest rates have been kept low for an unprecedented amount of time, it should be no surprise then that the downstream consequences too are unprecedented. Most obvious amongst these consequences has been the total fragility of the financial system, as highlighted by the recent collapses of Silicon Valley Bank, Signature Bank, and First Republic Bank, in rapid response to the semi-normalization in interest rates. This is only the tip of the iceberg, and the longer interest rates remain raised, the more the economic growth of the last decade will be shown to have severe structural problems. In the case of cloud computing, there are already signs that a half a trillion dollar industry that came to maturity entirely in this easy money time period is in trouble. Condensing the Cloud What do we mean when we talk about the “The Cloud”? For most, the public cloud is defined solely by the outsourcing of hardware management to a specialized provider. Despite the adage, the cloud isn’t just someone else’s computer. To be abundantly clear, there are incredibly good reasons for companies to outsource the physical management of compute to other market players better suited to the task. These decisions are rooted in the simple ideas of comparative advantage and economic specialization.&nbsp;However the major hyperscalers in the cloud oligopoly (AWS, GCP, Azure) share specific traits which define the entirety of their business models, modes of operation, and trade-offs that are far more nuanced than simply outsourcing hardware management. These massive platforms derive most of their usage, stickiness, and earnings not from compute alone, but ultimately from the additional services and tools tightly bundled with that compute. This includes a vast catalog ranging from managed databases, to identity management, CDNs, managed orchestration tools like Kubernetes, and so much more. In essence, the hyperscale cloud providers today are an intricate and often confoundingly complex array of proprietary tools and managed services. Why? To lock users into their platforms by bundling their software and services add-ons with their hardware. Want to switch to using a competitor’s compute resources instead? Not so fast: a cobweb of interlocking managed software dependencies now keeps you tethered to your host with no easy way out. This is the cloud today: a few major walled gardens defined by the bundling of hardware, software, and services in an effort to de-commoditize what is and should be a commodity business: supplying compute.&nbsp;The resulting size, concentration, and control over digital infrastructure that these providers leverage have subsequently become a far cry from the early visions of an open and decentralized internet. This state of affairs is a direct result of the macroeconomic incentives that have been in place almost continuously since 2008.&nbsp; All That Glitters isn’t Gold The deliberately “bundled” design of the cloud offers a clear tradeoff: speed over portability and flexibility. Lack of portability results in higher fiscal costs to customers, especially long term, as these providers know moving to a competitor is no trivial matter. The more services and tools customers use from their current cloud host, the more they are disadvantaged over time as the cost of migrating to a competitor tends to only increase. While using these more easily integrated tools and managed services to increase productivity and optimize for growth may have seemed like a straightforward choice to the customer it is ultimately a Faustian bargain. For most players operating in an economic environment defined by unprecedentedly low interest rates, this optimization for immediate growth over longer term considerations made total economic sense.In such a case, fundraising is easy, and as such fiscal discipline is not as valued by investors as speed to market and growth at all costs. If this sounds far fetched, believe me, it was extremely common up to early 2022 for angels and VCs to tell portfolio companies they weren’t spending money fast enough. When cash is cheap, other things are simply more important, and rationally so within the context of the economic environment. The costs of increasingly large AWS bills hardly registered as long as revenue growth continued and fundraising remained extremely easy, both themselves buoyed by low interest rates driving less discretionary spending all across the economy. If your competitors are leveraging the dynamics of the easy money environment, and you’re not, you’re likely dead in the water before the longer term considerations even matter. However unlike death or taxes, artificially low interest rates aren’t and can never be constant. As inflation concerns began to reach fever pitch over the last year, the Federal Reserve and other central banks have quickly reversed course by aggressively hiking interest rates once more. Very abruptly, the calculus driving economic decision making has turned on its head, and the impact on many industries, like banking, have been hard and obvious. Many other’s have been heavily impacted as well, albeit less dramatically. For now. The cloud included. Revenue and fundraising capabilities are both sinking in lockstep, and suddenly the abrupt need for aggressive fiscal discipline for the first time in over a decade is bringing to light certain glaringly obvious truths.&nbsp; The Cloud Doesn’t Actually Make Much Sense As it turns out, making yourself entirely reliant on a single provider with absolute leverage over your infrastructure is a really bad deal if you actually have to think long term or care about fiduciary responsibilities. That’s the conclusion many companies are rapidly coming to, a notable recent example being 37 Signals, as explained by co-founder David Hansson the creator of Ruby on Rails.In their effort to move from the Cloud, the team at 37Signals has built MRSK, a framework to help manage higher level tooling and application orchestration on top of raw compute, completely agnostic to where that compute is hosted. This serves a crucial function: It effectively helps unbundle some of the functionality typically bundled by cloud providers, giving you greater portability and control over your software. There is still a bit more additional work involved with this solution compared to ready-to-go cloud tooling to be sure, but MRSK and tools like it are relatively niche and underdeveloped, especially when compared to the vast investment the public cloud tooling has seen over the past decade plus.These tools and the paradigm it represents offers a starkly different alternative to how the cloud works today. By simply buying or renting raw compute and running the open source infrastructure stack of your choice on top of it, an “un-bundling” occurs which undoes the position of power the cloud oligopoly have created for themselves while the industry has slept. In this model, businesses can crucially still leverage more specialized third parties to manage their physical compute in the actual actual data center, and still leverage third party software to maximize their productivity, but in a portable and natively agnostic fashion. In our now higher interest rate environment, that reality of cloud costs and it’s long term implications for organizational flexibility and efficiency is becoming much less palatable for a huge swath of companies all at once. As more continue to flee the lock in and corresponding costs of the cloud out of pure fiscal necessity, the tools to make this “un-bundling” movement even easier and more practical will continue to be improved upon. With that in mind, we can scarcely imagine how the ecosystem of independent bare metal providers and portable infrastructure tooling might look like 10 years from now, or how they might have looked like today had there not been more than 10 years of artificially low interest rates to begin with. “The Cloud” as we know it today, may never have come to be. Indeed, the fact that it did evolve in the way it did increasingly seems like the anomaly. A development that was broadly accepted but is in fact seeming more counterintuitive in hindsight by the day. In it’s stead, we are seeing the growth of a much more diverse, distributed, and commoditized pool of compute providers, upon which an ecosystem of open source and highly portable software can flourish.&nbsp; Unwinding The Cloud The problem with the growth of major cloud providers isn’t the idea that certain businesses shouldn’t specialize in running the hardware on behalf of other businesses with other priorities. The problem is actually that that specialization doesn’t go far enough. There should be a competitive industry for offering compute as a commodity, and distinct industries for orchestration, managed services, and tool chains to effectively utilize that compute, independent of where the physical hardware is located and who manages it. Certainly, there will always be a market for vertically integrated hardware and software products, typically thought of as hosted platforms as a service. But if we accept that the expansive proliferation of this bundled pattern was in large part driven by unsustainable incentives derived from low interest rates, then we must conclude that the relative market for these offerings is much smaller than what we’ve seen up to this point. The logical conclusion of this is a continued “un-bundling” movement, and the gradual but accelerating transformation of a more than half a trillion dollar industry. That means growth for providers of "less-strings-attached" bare metal as a service (BMaaS) and basic virtualized offerings, and vastly increased interest, attention, and development of portable software and services that can be deployed on top of any provider of compute resources. This reorientation of how we build and deploy applications will have profound effects, but ultimately will likely be a huge blow to the defacto oligopoly and over centralization of infrastructure in the hands of AWS/GCP/Azure, and a big win for industry disintermediation and competition in the long run.In the midst of ongoing economic uncertainty, we might actually be seeing the internet trending towards a more natural and distributed equilibrium as skewed incentives are removed from the markets comprising it. One thing seems certain: Were it not for artificially low interest rates to begin with, the way we would use “someone else’s computer” would look very different from the cloud we see today, and much more in line with the original vision of an open, decentralized, and interoperable internet. Practical Takeaways For developers and dev ops engineers: Instead of that next AWS certification, start familiarizing yourself more seriously with bare metal, and how to deploy and work with more fundamental building blocks. Frameworks like MRSK, Supabase, and others offer paths to effectively deploy to bare metal as well as VPSs in manageable ways. Demand for technical talent that can facilitate migrations to this paradigm will continue to grow rapidly. For companies and organizations: Are you on the cloud? If so, you’ve likely already taken a recent look at your AWS bill and are wondering how to make some improvements. It’s time to start thinking outside of the box, and ask your engineering teams to take a serious look at some of these trends and technologies mentioned in this write up and present a plan to get to higher ground. Still operating on-premise infrastructure? You might have actually dodged a bullet there, and you’re in a position to carefully plan what your future looks like. There’s plenty of options between remaining on-prem, out sourcing on-prem management, going to a BMaaS model, or some hybrid of these options. For data centers and independent compute providers: This is your time to shine. There’s never been a bigger opportunity to grow your business, but to do so you need to offer comparable user experiences and programmatic capabilities to the large cloud providers to make transitions easier for new customers, but otherwise simply focus on delivering superior hardware for the price. At Hydrahost.com, we’re building a data center directory, and helping our data center partners offer modern Bare Metal Self Service capabilities with easy to use UIs and robust API. Contact us to learn more.

Discover how low interest rates accelerated cloud adoption and the challenges emerging as rates climb, reshaping digital infrastructure strategies.

The Cloud: A Low Interest Rate Phenomena 

In software, everyone wants to be in on the Next Big Thing™. Of course, identifying that Thing can be tricky. Today, Hydra is doing that work for you. In this primer we’ll cover: <ul><li>What is Sky Computing, and why it’s quickly emerging as the Next Big Thing</li><li>Sky Computing use cases, and how it compliments other trends such as AI/ML</li><li>How it works, the different implementation approaches, and their trade-offs</li><li>How industry adoption is likely to play out</li><li>The early companies and projects, like Hydra Host, poised to be early major players</li></ul> Whether you’re a software developer, platform/DevOps engineer, or a business owner, if you rely on the cloud and want to keep a competitive edge, being early to the Sky Computing discussion is a must. The goal of this article is to provide a high-level primer on this subject and its ongoing emergence, with actionable starting points on how to get in on the ground floor to the benefit of your career, business, and customers. <h2>What even is this Thing</h2> “Sky Computing” registers under 1,000 monthly searches on Google, so what is it, and why are we certain it’s not just a Thing, but poised to be a Big Thing? The term was cemented relatively recently in a paper titled <a href="https://arxiv.org/ftp/arxiv/papers/2205/2205.07147.pdf" rel="noopener noreferrer" target="_blank">“The Sky Above the Clouds: A Berkley View on the Future of Cloud Computing” by Ion Stoica and company</a>, and is meant to evoke a radical extension and enhancement of today’s cloud computing landscape. As the name suggests, the central idea of Sky Computing is users should be able to leverage all different clouds and other resource providers, whether major hyperscalers, independent datacenters, or edge compute, as easily as they’re able to leverage a single cloud provider. Rather than being reliant on any single cloud, the entire skyscape of computing should be at all of our fingertips. It’s worth noting that there are competing terms being used to describe the same ideas, including <a href="https://siliconangle.com/2022/05/21/supercloud-becoming-thing/" rel="noopener noreferrer" target="_blank">“Supercloud” and even “Metacloud”</a>. None have a clear edge yet, but Sky Computing has the benefit of being a completely novel term without any of the potential baggage of previous associations. The “Sky Above The Clouds” paper also does the best job of comprehensively describing the technology and challenges so far. <h2>The Problem</h2> Cloud was a pivotal step forward for the software industry. The cloud allowed a natural division between two compute spaces - the building and maintaining of applications and the operating hardware required to host them. However, the cloud as it works today also has significant down-sides. Today, all the major clouds have one anti-feature in common: cloud vendor lock in. Once you’ve built your entire infrastructure to a particular provider, shifting to another one entirely is typically a complicated and expensive undertaking. As a result, what we have today are different isolated “clouds”, or walled gardens with little agility and optionality for businesses. On the macro scale, this reduces competition and increases pricing. On the micro scale, this means the norm is to be forced to entirely rely on a single hosting operator for business operations, continuity, and their ongoing fair pricing and policies. <h2>Multi-Cloud is a Cry for Help</h2> Businesses already intuitively understand the criticality of the issues associated with being stuck with a single provider and are working to imperfectly address them. <a href="https://www.prnewswire.com/news-releases/virtana-research-finds-more-than-80-of-enterprises-have-a-multi-cloud-strategy-and-78-are-using-more-than-three-public-clouds-301548266.html" rel="noopener noreferrer" target="_blank">In a recent survey, more than 80% of enterprises have a multi-cloud strategy and 78% are already using more than 3 public clouds.</a> The factors pushing these enterprises towards leveraging new providers stretch from cost optimization, to available geographic locations, and to the need for diversification in the event of a provider outage. However, in Hydra’s experience of offering multi-cloud advisory and consultation services, most businesses are multi-cloud in the sense that they leverage other providers to deploy new services, rather than making legacy services cloud agnostic and multi-cloud capable. This is because migrating legacy infrastructure built to a specific cloud remains so difficult. Managing disparate resources across providers creates its own problems however, due to the burden of orchestrating and maintaining multiple different cloud infrastructure stacks and accounts which sport their own distinct tool chains and interfaces. There’s a better way, and that’s exactly what Sky Computing aims to achieve. <h2>The Next Big Thing</h2> Imagine a single dashboard that allows you to click and choose which provider you deploy your resources to, filterable by variables such as prices, locations, latency estimates, historical uptime, performance, compliance standards, and legal jurisdictions. You can deploy any of your services and applications independently to whichever provider is a superior fit for the needs of that particular component, while also redeploying them as often as there is new favorable options to take advantage of. Substituting this dashboard with an API or Terraform provider enables the same capabilities programatically, allowing you to build pipelines which can deploy and redeploy services to different providers automatically depending on stated preferences. One obvious use case for these capabilities is the ability to shop around for the best deal on compute intensive operations, such as model training for machine learning. Whenever you’re ready to begin a new job training a model from a large data set, you can do an active search of the least expensive (or most performant) hardware available on the market to leverage. Another use case is the ability to compare and choose between all resources available that meet required jurisdictional or compliance guidelines, such as GDPR, HIPAA, or ISO. Yet another can be optimizing for uptime, by spreading clusters of the same critical service across multiple different providers as an active risk management strategy. <h2>How It Works</h2> The key component required for unlocking this experience and the subsequent use cases is described by the Sky Above the Clouds authors as a “intercloud brokerage”, or what Hydra Host originally conceived as an “arbitrage layer.” These brokers serve as a middleware layer and API gateway for each cloud provider, providing a unified and standardized interface for leveraging all of their resources. They do the work of integrating and maintaining the interfaces of the various cloud providers, so you can focus on your application level services. Several projects are already pursuing differing visions of this broker-like model today, Hydra Host included. Differing approaches exist with different sets of tradeoffs. The largest distinction to be made in categorizing them so far is what we call either the maximalist approach or the minimalist approach (what the Sky Above the Clouds refers to as a “shim layer”). <h2>Maximalist vs Minimalist</h2> The goal of the maximalist cloud broker approach is to piece together and standardize the interfaces for the various product catalogues of the major cloud service providers. This can include high level services such as EKS, RDS, data analytics tools, and more. The advantage here is the significant demand and use cases for combining the best services from all providers, say RDS from AWS and data analytic tools from GCP, as they are widely used. Given the sheer number of services that cloud providers now provide, and the inherent complexity of higher-level services, there is a disadvantage. Building and maintaining an ever-growing list of products across clouds becomes tricky to scale and maintain even for a company dedicated to the function. On the other hand, the goal of the minimalist approach is to conversely bridge only the lowest levels of cloud and hosting resources together, such as compute, storage, and networking. Once bridged, a protocol like standard is provided on top so higher-level services can be rebuilt on to a common interface. This interface is natively cloud agnostic, meaning applications built on top of it are able to run uniformly in all hosting environments without changes. The advantage to the minimalist approach is that integrations with different providers are far simpler to build and maintain than the maximalist approach. There are fewer moving targets and less complexity. Like any good protocol, this approach makes the “broker” strategy much simpler to understand and lightweight to run. Additionally, it is easier to quickly integrate hosting providers. This expands possibilities far beyond the major cloud providers, including the large landscape of independent data centers and rapidly growing number of edge compute locations. Furthermore, this approach forces providers to compete on the commodity level that is the hardware, increasing competition where it can have the biggest daily impact: your wallet. The downside of the minimalist approach is development complexity. Development shifts from integration work to redeveloping the infrastructure stack, meaning that the platform won’t be ideal right away for every application if they are heavily reliant on higher level cloud services like EKS. This will be overcome through time as the infrastructure stack matures and offers effective analogues to common high level cloud services. <h2>Adoption Roadmap</h2> So, what will the rollout of Sky Computing look like? Right now, adoption is nascent, which provides more upside for the innovators and early adopters. Ultimately, the adoption path will be influenced by one major factor: friction. While cloud brokers make it easy to deploy applications to multiple clouds by offering a unified interface, migrating off cloud interfaces to sky brokers takes the same type of work that makes cloud migration a pain to begin with. This means that similar to current multi-cloud trends, most early adoption of the intercloud broker model will come from new services and applications. Emerging startups and SMBs are uniquely positioned to leverage this technology, as they are relatively unencumbered by sprawling legacy applications. However, another group that that is well positioned to be early adopters includes large enterprises who are already facing the need to undergo infrastructure migrations. Perhaps this is due to recent M&amp;A or simply an effort to dramatically cut costs. For comparable effort, these enterprises can migrate to a Sky Computing model to solve their dilemma and never have to worry about extensive hosting migration pains ever again. <h2>Getting Involved</h2> It’s not often that you’re able to get in on the ground floor of a powerful new paradigm, but here we are. There are multiple “broker” providers to explore which we’ve outlined here (non-exhaustively, we’re sure), and given the early stages all are easy to reach out to. We, of course, encourage you to read the original <a href="https://arxiv.org/ftp/arxiv/papers/2205/2205.07147.pdf" rel="noopener noreferrer" target="_blank">"The Sky Above the Clouds"</a> paper, and other links in this article. Here at Hydra Host we’re working on launching a State of the Sky newsletter you can subscribe to if interested, as well as a community chat which we’ll keep subscribers alerted to. Sky Computing is poised to effectively address rampant pain points in today’s cloud computing landscape and change the entire cost benefit equation distributed computing. Together, we can build a community of early practitioners that can accelerate this change and provide outsized benefits for early adopters. <h2>Inter-Cloud Brokers:</h2> <ul><li>Hydra Host: We’ve been researching and experimenting with various approaches to the “arbitrage/broker” function since April 2021, settling on a Minimalist strategy. We’re working with early adopters to onboard them on to our fully managed platform, with plans to open source our software in the second half of 2023. 2023 roadmap goes beyond public clouds, and integrating independent bare metal value providers. Make sure to contact us for a demo or if interested in our future newsletter, and keep an eye on our blog!</li><li><a href="https://controlplane.com/" rel="noopener noreferrer" target="_blank">Controlplane</a>: Founded in 2019, it allows user to mix and match cloud services from the big 3 (AWS, GCP, Azure) in the vein of our described “Maximalist” Sky strategy.&nbsp;</li><li><a href="https://www.upbound.io/" rel="noopener noreferrer" target="_blank">Upbound</a>: Founded in 2017, Unbound is building a universal cloud platform powered by their crossplane framework, which can be self hosted to help manage a variety of service providers.&nbsp;</li><li><a href="https://www.anyscale.com/" rel="noopener noreferrer" target="_blank">Anyscale</a>: Founded in 2019, Anyscale is a platform to run “Ray” based AI/ML workloads on top of any cloud. From the outside they seem to employ a more minimalist strategy to ensure workloads are inherently more portable, and is focused entirely on the hosted Ray use case, for now.</li></ul> We’ll be expanding this list on an ongoing basis as we see more players in the inter-cloud broker space continue to appear. If we missed anyone, let us know!

Explore Sky Computing's potential to unify cloud services. Get seamless integration and enhanced flexibility beyond traditional cloud limitations.