Should you invest in CoreWeave’s IPO?
This morning, CoreWeave rang the Nasdaq bell, becoming the first dedicated AI company to go public in the post-ChatGPT era.
The company ultimately raised ~$1.5 billion - well below its initial target of $2.5 to $4 billion - and priced meaningfully lower than the $47–$55 range originally floated by bankers. Still, as the first "pure play" AI company to go public post ChatGPT, CoreWeave's listing will set a precedent for other AI companies eyeing the public market.
But beyond the headlines and hype, how should investors evaluate this opportunity? In the analysis below, I dive into CoreWeave’s S-1 and draw insights from conversations with industry experts and institutional investors. The goal is to clearly understand CoreWeave’s fundamentals, and to evaluate the strength and risk of the company.
A quick history of CoreWeave
CoreWeave was founded in 2017 originally as an Ethereum crypto mining venture called Atlantic Crypto. Following the 2018 crypto market downturn, the founders (who were all former natural gas commodities traders) rebranded the company to CoreWeave and began diversifying beyond cryptocurrency. By 2022, the company had fully transitioned to providing cloud-based GPU infrastructure tailored for AI workloads.
The company's strategic pivot attracted significant investor interest, leading to substantial funding rounds. Since 2023, the company has raised more than $14b in debt and equity across 12 financing rounds, including a massive $7b round debt round last year from Blackstone. Revenue has growing impressively, skyrocketing from $16m in 2022 to $1.9b in a two-year span.
The Bull Case
CoreWeave presents a compelling opportunity to invest in the infrastructure layer of what may be the most transformative technology cycle in decades.
Generative AI is still in its early innings. While adoption is accelerating in areas like software development and customer support, most use cases remain nascent. At the same time, the industry is undergoing a pivotal shift—from training-heavy compute to test-time inference—which will only amplify the need for high-performance, scalable infrastructure.
CoreWeave is uniquely positioned to meet that demand. In just three years, the company has:
Built a fleet of 32 data centers
Deployed over 250k GPUs
Built an end-to-end data center stack, including a robust software inference platform and tooling, bolstered by the recent acquisition of Weights & Biases
The company’s financial growth has been impressive:
Revenue grew by more than 100x in two years, reaching $1.9b in 2024
During the same time period, Adjusted EBITDA grew from negative $10m in 2022 to $1.2b
While the company’s balance sheet is highly levered, two factors significantly reduce risk:
Strong unit economics, with an average cash payback period of just 2.5 years
A “just-in-time” financing model that provides capital flexibility even in a highly CapEx-intensive environment
If the team can continue executing while managing an appropriate capital structure, CoreWeave could emerge as a long-term leader in AI infrastructure.
1. Market structure and tailwinds
We are still early in the infrastructure buildout for AI. History shows that major technological shifts – like the railroads in the 19th century or the internet and telecom boom in the late 20th –require years of sustained infrastructure investment. During the dot-com era, more than $500 billion (over $1 trillion in today’s dollars) was poured into building out networks and data infrastructure, and it took nearly a decade for that investment cycle to peak. By comparison, we are only in the third year in the GenAI buildout. Total CapEx spent has been estimated in the $200-300b range, large but still only a fraction of the internet buildout. Jensen Huang, in the latest GTC keynote address, expects data center build-out to reach over a trillion-dollars before long. The Project Stargate project alone will see investments of up to $500b in the US by 2029.
Furthermore, it appears that the cloud market is getting increasingly fragmented. CoreWeave is the poster child of this, but there are other “neo-clouds” including Lambda Labs, Nebius, TensorWave (Translink investment). Marketplace players have popped up recently as well such as Lepton AI, HydraHost and Shadeform (another Translink investment). Pitchbook estimates the “neo-cloud” market to grow 8x from $4b to $32b by 2027. (link)
A key driver behind this fragmentation is the strategic repositioning of chipmakers such as Nvidia. These chip manufacturers are actively working to diversify their customer base and reduce dependence on the hyperscalers, who are developing proprietary chips themselves.
Simultaneously, enterprises seeking specialized computational infrastructure tailored for specific AI training and inference use cases may favor these neo-clouds and specialized marketplaces. These platforms can offer greater flexibility, customization, and potentially lower costs compared to traditional hyperscalers.
For both these reasons, further fragmentation in the AI cloud market appears likely.
2. End-to-end AI data center
AI-specialized data centers such as CoreWeave differ significantly from traditional cloud data centers. While conventional cloud data centers are built for general-purpose workloads—search, e-commerce, databases—using a blend of CPU and GPU resources, CoreWeave is purpose-built for AI.
Its infrastructure is optimized for training and inference, featuring high-density GPU clusters, high-speed interconnects (e.g., InfiniBand, NVLink), and liquid cooling. These infrastructure design choices enable lower latency, higher throughput, and more flexible resource configurations than general-purpose clouds.
Based on conversations, CoreWeave’s reliability is reportedly in the high-90% range, which is well above that of the hyperscalers. As a result, CoreWeave cloud can offer up to 20% improvement in system MFU (Model FLOPS Utilization). Ultimately, this translates into better performance at lower cost.
Beyond hardware, CoreWeave has built a vertically integrated software stack to simplify infrastructure provisioning and workload orchestration. Its recent acquisition of Weights & Biases further extends this platform, making it an end-to-end platform that serves both infra developers and app developers that are looking to build, fine-tune, or experiment with AI models.
Strategically, this acquisition enables CoreWeave to compete more directly with hyperscalers like AWS and Google, whose cloud business bundles with DevOps tooling (e.g., SageMaker, Vertex AI). It also diversifies CoreWeave’s business model from a pure compute leasing model into SaaS and subscription revenue. From that angle, the W&B deal is strategically significant.
3. Access to power and chips
Going forward, the real limiting factor in AI may not be chips but the electricity needed to power these chips. CoreWeave anticipated this early and secured long-term power contracts. This should provide sufficient headroom to meet medium-term compute demand.
"We ran into a set of constraints, which are everything because data centers don't get built overnight. So there is DCs. There is power. And so that's sort of been the short-term constraint."- Satya Nadella, CEO of Microsoft
In just two years, CoreWeave has emerged as one of Nvidia’s most important global partners, with access to over 250,000 H100s. This relationship is expected to deepen further, especially given Nvidia already owns 6% of the company and is set to be an anchor investor in the upcoming IPO.
4. Strong top-line growth and unit economics
CoreWeave’s revenue has grown 7x over the past year and is projected to more than double again to $4.6b this year. These forecasts are underpinned by the existing $15b in remaining performance obligation (akin to “booked revenue” in SaaS parlance) of which 54% or approximately $8b will be recognized in the next two years. This figure does not include the recently signed $12b, 5-year contract with OpenAI, which should suggest meaningful upside over current forecasts.
Furthermore, 96% of CoreWeave’s contracts are long-term commitments with an average contract length of 4 years, which should lead to high revenue visibility. Unit economics are also strong: the average cash payback period is 2.5 years, meaning the company becomes cash flow positive slightly beyond the halfway mark of its typical contract. Given the implied cash ROI is 40% (1/2.5 years) and the cost of capital is only 10-15%, the company is operating with highly attractive unit economics.
Quote: As of December 31, 2024, our committed contracts had a weighted-average contract duration of approximately four years…we anticipate that our average cash payback period, including prepayments from customers, will be approximately 2.5 years. Our cash payback period is the time we anticipate it would take to break-even on our investment in GPUs and other property and equipment through adjusted EBITDA. - CoreWeave S-1
5. Flexible debt arrangements
Some pundits have drawn comparisons between CoreWeave and WeWork, but the analogy breaks down under further scrutiny. The two models are fundamentally different.
WeWork’s collapse was driven in part by a structural mismatch: long-term lease liabilities backed by short-term rental revenue. CoreWeave, by contrast, operates on a just-in-time funding model. It doesn’t issue GPU purchase orders unless backed by long-term committed customer contracts. This means debt is only incurred once revenue is secured, which aligns liability duration with revenue duration.
We typically do not submit purchase orders for systems without having a committed contract that matches the level of compute generated by such systems. We generally submit purchase orders for systems on a just-in-time basis at the time of contract signing. This ensures that, generally, we provide infrastructure to customers as needed, versus making large investments without certainty of payback. – CoreWeave S-1
CoreWeave also uses Delayed Draw Term Loans (DDTLs), which are collateralized against contracted cash flows and infrastructure assets. This structure not only mitigates default risk but also lowers the company’s overall cost of capital.
In summary…
Taken together, CoreWeave offers a rare combination of hypergrowth, strong unit economics, and strategic positioning in one of the most capital-constrained parts of the AI stack. Its just-in-time financing model and long-term customer contracts de-risk execution, while access to chips, power, and talent provide defensibility. As the AI infrastructure market fragments and scales, CoreWeave will become a key player in this space.
The Bear Case
Behind CoreWeave’s rapid growth and scale lies a more complex and nuanced risk profile. The main risk of the company include: (1) a small set of highly concentrated customers (2) a leveraged capital structure, and (3) governance and internal controls that warrant closer scrutiny. While none of these issues in isolation may be a dealbreaker, taken together, they present a risk profile that’s difficult to underwrite for most investors.
1. Customer concentration
The most obvious risk is customer concentration. CoreWeave’s largest customer, Microsoft, accounted for 62% of their revenue in 2024. This gives Microsoft significant negotiating leverage when deciding on terms. Already, there has been reports that Microsoft has canceled some contracts.
The customer concentration may be reduced by the recently announced $12 billion contract from OpenAI, but that comes with its own risks:
OpenAI is still a startup (albeit, a large startup), and its solvency is not guaranteed. With the looming threat of Deepseek and other open-source alternatives, OpenAI’s success is far from guaranteed.
The interest rate CoreWeave pays is tied to quality of the customer. Investment grade customers like Microsoft allow CoreWeave to borrow at much lower rates with spreads of 6.5% over SOFR (secured overnight financing rate). By shifting its customer base to less mature players, its overall financing costs would rise significantly
2. Leverage
CoreWeave’s business is inherently capital-intensive. While customer prepayments provide some upfront relief, the company still faces significant cash outlays. In theory, this model is fine as long as business fundamentals remain strong and financiers are willing to roll over the debt. But as both leverage and the cost of capital rise, so does the company's financial risk profile.
Today, CoreWeave is carrying approximately $8 billion in total debt, with a blended interest rate of 11% - implying nearly $1 billion in interest payments due this year alone. Additionally, principal repayments are also accelerating over $2.5 billion in principal is due this year, bringing the total cash obligation for debt service (interest + principal) to more than $3.5 billion in 2024.
In addition to its debt load, CoreWeave faces $15 billion in future undiscounted lease commitments, primarily for data center and office buildouts. These leases - set to commence between 2025 and 2026 - will result in further liability for the company. Therefore, the total indebtedness of CoreWeave including the lease obligations is closer to $23 billion.
“As of December 31, 2024, the Company executed additional lease agreements, primarily for data centers and office buildings, that had not yet commenced. The aggregate amount of estimated future undiscounted lease payments associated with such leases is $15.0 billion. These leases will commence between 2025 and 2026 with estimated lease terms of five to sixteen years.” – CoreWeave S-1
The implication here is clear: the majority of CoreWeave’s operating cash flow will be consumed by servicing debt and lease obligations. To continue growing at its current trajectory, the company will need to raise additional capital - likely through further debt issuance at potentially higher interest rates. That’s fine if demand and prices of compute continue to hold steady. But if there’s a reversal, the deleveraging may be particularly painful for CoreWeave
Finally, just as debt is required to grow the business, debt covenants may begin to constrain future growth. The company is required to maintain a net leverage ratio of 6.0x. While these covenants are prudent and appropriate, it may limit flexibility in a situation where continued growth remains heavily dependent on capital availability.
3. Governance and internal controls
While customer concentration and leverage are critical risks, arguably the most concerning vulnerabilities lie on the governance side.
First, there is a question of founder alignment. Just months before filing, the founders reportedly sold $500 million worth of shares to Fidelity in a secondary transaction. While founder liquidity is not inherently problematic, such a sizable cash-out ahead of IPO may raise questions about the long-term commitment of the founders.
Second, CoreWeave has adopted a super-voting structure. Despite owning less than 30% of the economic interest, the co-founders collectively control over 80% of the voting power. While not uncommon amongst tech companies, this governance design significantly limits the influence of outside investors and entitles the founding team to maintain control over all major decisions.
Most troubling, however, are the disclosures related to internal control deficiencies—particularly within the finance function. The S-1 filing cites several material weaknesses in internal control over financial reporting, including:
Lack of effectively designed and maintained IT general controls over systems that support financial reporting
Insufficient segregation of duties across finance-related functions
Inadequate number of qualified personnel in the accounting and operations teams with the appropriate expertise to ensure accurate transaction recording and disclosure
“We have identified material weaknesses in our internal control over financial reporting. If our remediation of such material weaknesses is not effective... our ability to produce timely and accurate financial statements or comply with applicable laws and regulations could be impaired.” – CoreWeave S-1
Taken together, these issues suggest that CoreWeave may still be in the process of developing the institutional maturity, governance practices, and operational rigor typically expected of a company approaching the public markets. While none of these challenges are insurmountable, they highlight areas that will require more scrutiny from investors.
In summary…
While CoreWeave is undoubtedly an emerging player in the AI infrastructure ecosystem, it is doing so on an overly ambitious and fragile foundation. The company’s dependence on a small number of customers, significant financial leverage, and immature internal controls create meaningful risks.
Conclusion
CoreWeave represents the classic high-risk/high-reward investment.
On one hand, the company is among the fastest-growing players in a once-in-a-generation infrastructure shift. While some have argued GPU demand has peaked, conversations with industry participants suggest demand is still healthy. Stargate alone will be a $500 billion investment. This creates a favorable market environment for AI infrastructure providers, at least in the short-to-medium term.
On the other hand, CoreWeave’s aggressive growth is underpinned by significant financial leverage, and continued growth depends on ability to continue refinance debt as they mature. Simply put, CoreWeave’s biggest strength (hypergrowth) is also its biggest risk (high leverage).
At an enterprise value of roughly $30 billion, CoreWeave trades at around 10x forward EV/EBITDA—a valuation that looks reasonable relative to its risk profile. With few publicly traded pure-play AI infrastructure companies available, CoreWeave presents another option for investors seeking direct exposure to the AI thesis.
Ultimately, the decision to invest will depend heavily on individual risk appetite.
But one thing’s for sure – expect a ton of volatility.