Startups are reshaping the tech industry by addressing the growing demand for GPUs and challenging the dominance of giants like AWS and Google Cloud.
In 2014, Vultr founder David Aninowsky focused on building data centers for traditional cloud computing, powered primarily by CPUs. GPUs, then associated mostly with gaming, were hardly on corporate radar. Fast-forward to today, GPUs are critical for AI model training, driving Vultr to a valuation of $3.5 billion. This Florida-based startup has raised $333 million from AMD and LuminArx Capital this month alone, with GPU rentals now fueling its rapid growth. “AI is the fastest-growing segment in infrastructure,” says Vultr CEO JJ Kardwell.
Vultr isn’t alone. Investors have poured roughly $20 billion over the past year into 25 companies renting GPU access, according to Forbes’ analysis of Pitchbook data and corporate filings. This includes $8 billion in equity and $12 billion in debt from backers like BlackRock and Pimco.
A Rapidly Growing Sector
The emerging sector, often dubbed “NeoClouds,” is led by companies like Coreweave. Initially a cryptocurrency miner, Coreweave pivoted to renting GPUs for AI after the crypto crash in 2018. Over the past year, it raised $1.75 billion in equity and $8.1 billion in debt, reaching a $23 billion valuation.
Other players include former crypto startups like Crusoe Energy and Germany’s Northern Data, alongside established data center firms like Vultr and France’s OVH. Nebius, spun out of Yandex, has raised $700 million this month to expand its GPU rental operations.
Even smaller players like Runpod and Fluidstack are gaining traction by aggregating idle Nvidia chips and selling them at competitive rates.
Competitive Pricing Drives Growth
NeoClouds have thrived by offering significantly lower prices than major cloud providers. For instance, Coreweave charges $2.21 per hour for Nvidia A100 GPUs, while AWS charges $5.12 per hour. The difference? NeoClouds rent out bare-metal GPUs, forgoing the software integration that big players like AWS provide. This makes them particularly appealing to startups focused solely on cost-efficiency.
However, this lean approach isn’t without risks. Building and managing sophisticated data centers is far more complex than mining cryptocurrencies. “It’s like comparing an industrial warehouse to a luxury hotel,” warns Crusoe Energy CEO Chase Lochmiller.
Scaling Rapidly to Meet Demand
NeoClouds have shown remarkable agility. Coreweave doubled its data centers in a year, now operating 28 sites, and secured a $10 billion, five-year contract with Microsoft. Similarly, Crusoe Energy set up a renewable-powered data center in under a year, far faster than traditional hyperscalers.
Major tech firms like Microsoft and Oracle are also turning to NeoClouds to supplement their capacity. Oracle, for example, reportedly signed a $3.4 billion deal with Crusoe Energy for a GPU-powered data center in Texas.
Challenges Loom Ahead
Despite their rapid rise, NeoClouds face significant challenges. Building and operating data centers requires massive capital and expertise. Additionally, a potential oversupply of GPUs and falling rental prices could squeeze margins. Some startups have already slashed prices by a third since September.
The arrival of new, more efficient Nvidia chips could further complicate matters, favoring larger, well-funded NeoClouds while pushing smaller players out of the market. “Many of these startups are unlikely to survive long-term,” predicts Dylan Patel, an analyst at SemiAnalysis.
As the AI market evolves, NeoClouds must navigate a delicate balance of scaling fast enough to meet demand while maintaining financial stability. Whether these startups become the next tech giants or fade into obscurity remains to be seen.
see also: what is artificial intelligence