Alternative clouds are booming as companies seek cheaper access to GPUs | TechCrunch

The desire for alternative clouds has never been greater.

case in point: corewaveThe GPU infrastructure provider, which started life as a cryptocurrency mining operation, raised $1.1 billion in new funding this week from investors including Coatue, Fidelity and Altimeter Capital. The round brings its valuation to $19 billion post-money, and increases its total debt and equity to $5 billion – a remarkable figure for a company less than ten years old.

It’s not just CoreWeave.

Lambda Labs, which also offers a range of cloud-hosted GPU instances, secured a “special purpose financing vehicle” of up to $500 million in early April, a few months after closing a $320 million Series C round. Voltage Park, a non-profit backed by crypto billionaire Jed McCaleb last October announced It is investing $500 million in GPU-supported data centers. And Together A.I.A cloud GPU host that also conducts generative AI research, raised $106 million in a round led by Salesforce in March.

So why is there so much excitement and money being poured into the alternative cloud space?

The answer, as you might expect, is Generative AI.

As the generative AI boom times continue, the demand for hardware to run and train generative AI models at scale is also increasing. GPUs, architecturally, are the logical choice for training, fine-tuning, and running models because they contain thousands of cores that can work in parallel to execute the linear algebra equations that make up the generative models.

But installing a GPU is expensive. That’s why most developers and organizations turn to the cloud instead.

The incumbents in the cloud computing space – Amazon Web Services (AWS), Google Cloud, and Microsoft Azure – offer no shortage of GPUs and specialized hardware instances optimized for generic AI workloads. But at least for some models and projects, alternative clouds may be cheaper – and provide better availability.

On CoreWeave, renting an Nvidia A100 40GB – a popular choice for model training and inference – costs $2.39 per hour, which works out to $1,200 per month. On Azure, the same GPU costs $3.40 per hour, or $2,482 per month; On Google Cloud, it’s $3.67 per hour, or $2,682 per month.

Given that generic AI workloads are typically performed on clusters of GPUs, the cost delta increases rapidly.

“Companies like CoreView participate in a market of what we call specialized ‘GPU as a Service’ cloud providers,” Sid Nag, vice president of cloud services and technologies at Gartner, told TechCrunch. “Given the high demand for GPUs, they provide an option for hyperscalers, where they have taken Nvidia GPUs and provided another route to market and access to those GPUs.”

Nag points out that some big tech companies have also started relying on alternative cloud providers as they face compute capacity challenges.

Last June, CNBC informed of Microsoft had signed a multi-billion dollar deal with CoreWeave to ensure that OpenAI, the creator of ChatGPT and a close Microsoft partner, would have enough compute power to train its generative AI models. Nvidia, purveyor of the bulk of CoreWeave’s chips, sees this as a desirable trend, perhaps for leverage reasons; It is said that this has given some alternative cloud providers preferential access For its GPU.

Lee Sustar, principal analyst at Forrester, thinks cloud vendors like CoreWave are succeeding partly because they don’t have the infrastructure “baggage” that existing providers have to deal with.

“Given hyperscaler dominance in the overall public cloud market, which demands huge investments in a series of infrastructure and services that deliver little or no revenue, challengers like Coreview will have the upper hand without the hyperscaler-level burden.” There is an opportunity to succeed by focusing on premium AI services. Overall investment,” he said.

But is this growth sustainable?

Sustar has his doubts. He believes the expansion of alternative cloud providers will depend on whether they can continue to bring GPUs online in high volumes, and offer them at competitively low prices.

Competing on pricing may be challenging in the future as companies like Google, Microsoft, and AWS increase investment in custom hardware to run and train models. Google offers this tpu, Microsoft recently unveiled two custom chips, Azure Maia and Azure Cobalt, And AWS has Tranium, Inferentia and Graviton,

“Hyperscalers will leverage their custom silicon to reduce their dependence on Nvidia, while Nvidia will focus on CoreWave and other GPU-centric AI clouds,” Sustar said.

Then there’s the fact that, while many generative AI workloads run best on GPUs, not all workloads need them – especially if they’re not time-sensitive. CPUs can run the necessary calculations, but are generally slower than GPUs and custom hardware.

More existentially, there is a danger that the generic AI bubble will burst, leaving providers with a glut of GPUs and not nearly enough customers demanding them. But Sustar and Nag say the future looks bright in the short term, with both upstarts expecting a steady stream of cloud deployments.

“GPU-oriented cloud startups will [incumbents] There is a lot of competition, especially among customers who are already multi-cloud and can handle the complexity of management, security, risk and compliance across multiple clouds,” Sustar said. ,These types of cloud customers are comfortable trying new AI clouds if they have trusted leadership, solid financial backing, and GPUs with no latency.