Understanding the Allure of Neoclouds, the Rainmaking AI Cloud Firms with Multi-Billion-Dollar Forecasts

Beyond generative chatbot use cases, Artificial Intelligence (AI) has become a widespread technology in the enterprise sector across verticals, bringing about hundreds of billions in investment over the past few years. With this, the infrastructure to support such services has become more in-demand, and Graphics Processing Units (GPUs) have become a costly commodity—giving rise to alternative cloud providers.

In its February Series D equity round, San Francisco-based neocloud provider Lambda Labs raised US$480 million, bringing up its post-money valuation to US$2.5 billion. GPU giant NVIDIA, as well as Supermicro, were among the investors involved within the fundraiser. Last year, it saw US$320 million in its Series C, and across rounds, the firm has raised over US$1 billion. The company was founded in 2012, as a response to the founders being frustrated with the costs of Amazon Web Services (AWS), which they had used in previous projects.

Similar to hyperscalers like AWS, Microsoft, and Google Cloud, neocloud firms, also known as GPU clouds, such as Lambda offer cloud and software services, which are used to train, fine-tune, and deploy AI models. However, unlike the giant, these are rarely hyperscalers. Indeed, Lambda has seen massive interest develop in its solution, but it isn’t the only one amassing such attention—the ecosystem is flourishing, given the AI demand.

Perhaps the largest player is CoreWeave, which raised US$1.1 billion in May 2024, and is reportedly looking to float its shares in an Initial Public Offering (IPO) at US$35 billion, which would potentially raise US$4 billion for the firm. The company has also grown from 14 to 28 data centers in the last year, with Microsoft reportedly signing a US$10 billion deal to run its AI models using CoreWeave’s infrastructure. Another formidable player is Crusoe, which raised US$600 million in its Series D in December 2024. Vultr, an “alternative hyperscaler” that operates across 32 cloud data center regions worldwide also raised a funding round in the same month, collecting US$333 million, and hitting a valuation of US$3.5 billion. Although there may be slight differences between competitors, reportedly over US$60 billion was funneled into these types of firms in 2024.

It may be inferred that hyperscalers would uphold a price advantage over these companies, given the market logic of economies of scale. After all, they have the most reach, the most complete solutions, and the greatest number of resources to support this scale. However, this is not the case, as neocloud firms have logged up to 90% relatively lower costs for their services. Further, hyperscalers are also engaging with these newer neocloud providers. GPUs, especially those from NVIDIA (which has invested in these players), are a limited resource. Through these providers, hyperscalers can support their growing AI use cases and capacity constraints, creating a symbiotic relationship.

Additionally, the reason that neocloud firms can remain competitive is that several mainly provision stripped-back, bare metal GPUs without the software, services, and ecosystem installation that firms like AWS push with their GPU-as-a-service (GPUaaS) offerings. In other words, they offer a leaner, no-frills product that allows them to keep prices down via lower Operational Expenditure (OPEX). Although this also brings about benefits for larger enterprises, it also opens up avenues for use by smaller firms, such as startups that only require access to the GPU technology. However, pricing is an obvious benefit of neocloud services, but it is far from being the only one.

Neoclouds offer less vendor lock-in, with easier integration as smaller firms are less likely to have sufficient operating revenue, especially when bootstrapped. Therefore, this will require integrating several unrelated and largely open-source systems. Neoclouds can provide the GPU component without locking enterprises into their whole ecosystem, as a hyperscaler like AWS would. This makes for greater flexibility and experimentation options in the beginning of the product development phase, as well as unlimited scaling options. Additionally, neoclouds offer a different level of security. As these firms are relatively smaller scale compared to their hyperscaler counterparts, they will be less of a target for cyberattacks, and have better, consolidated management of their resources, meaning that these firms are likely to be more secure. Players like Vultr that have a similar geographic reach to their hyperscaler counterparts also alternatively have more control over their servers, with a more proactive approach to security, and a more hands-on approach. Lastly, where hyperscalers offer a breadth of services as one-stop shops, neocloud firms focus their resources on their domains of expertise. Lambda Labs, for example, offers full-stack, self-service vertical solutions for enterprise AI models, which means that users can utilize the firm’s solutions independently without the need to tap other firms for components at a costlier price tag. Vultr, which has boasted some of the most low-cost solutions on the market, also offers 32 use cases across six primary industry verticals, leveraging its expertise in AI and data management and partnerships with AMD and NVIDIA. 

Although neocloud providers may also be cutting into the profits of some hyperscalers, the high demand for their GPU-fueled solutions have led to hyperscalers viewing these players as partners, not competition. Hyperscalers will still retain the large contracts for cloud storage and applications for multi-national companies, which differentiate them from a lot of these neoclouds—they won’t be going out of business anytime soon. These players are currently existing in harmony, but things could soon change as hyperscalers expand to offer more resources for greater longevity in the market. New GPUaaS entrants, such as SK Telecom, further threaten to cut into players’ market share.  It will be interesting to observe how these big players react to the different demands in the market as new technologies emerge.

It is clear that these firms are gaining billions in funding due to the unprecedented demand for AI services. However, when the hype equalizes, and GPU demand slows, these firms will struggle to keep afloat if the only solutions they offer remain specific to AI and GPUaaS. Eventually, firms will need to diversify their product portfolio and focus on either verticals or a variety of horizontal, yet niche solutions for enterprises across sectors. Naturally, neoclouds should continue to offer GPUaaS and AI-related services, even after the mass interest dies down. However, even as this hype continues on currently, neoclouds should already be building compatible use case applications such as data analytics, chatbots, or image processing to make their offerings more accessible to more customers, lowering the barrier of entry to neocloud services, and solidify their name in the cloud computing space.

The changing cloud ecosystem should not be seen as a dampener on hyperscaler activity—such firms should remain optimistic and partner with neoclouds to extend their AI services. Understandably, the competitive pricing that these smaller players offer may not be reasonably replicated by hyperscalers. However, these firms should engage neoclouds in their partner ecosystem, particularly in their AI tech stack. Neoclouds largely rely on open standards, providing hyperscalers with a chance to offer multi-cloud solutions. Hyperscalers must play to their strengths, offering targeted use cases and storage services, with more of the back end services, such as GPUaaS, being sourced from neocloud providers. As players such as Google Cloud, Microsoft Azure, and AWS have industry-standard, mature enterprise services, while most neocloud companies are more developer-centric, it may be wise to forge partnerships to play to the strengths of both types of companies.

From the enterprise side, neoclouds offer a variety of benefits for specific use cases. Primarily, neoclouds are beneficial for technology firms that employ staff with high developer expertise and provide or use AI as one of their core offerings or tools, given the largely self-service, stripped-back services these firms deliver. This is especially true for medium-sized firms that are only based within one country and do not require global services, as this allows for significant cost savings, provisioning resources only when and where they are needed. Of course, each neocloud firm is different, with distinctive value propositions. However, neocloud’s main differentiator from hyperscaler options are their lean, open solutions. Neocloud solutions would also best benefit startups that require quick, easy resources for prototyping AI projects without the need for an application ecosystem.