Mark Cuban recently warned that Nvidia is playing the role of a dot-com-era IPO machine, funding AI companies across the sector.
The concern centers on specialized cloud operators like CoreWeave and Nebius, which are spending billions on infrastructure (GPUs and data centers) before customer payments arrive—a timing gap that leaves them dependent on continued outside funding.
While both companies have substantial revenue backlogs, their near-term capital needs exceed their operating cash flow, creating vulnerability if financing conditions tighten.
What happened
Mark Cuban compared Nvidia's role in financing AI companies to 1990s IPOs that funded internet startups. Nvidia holds $42.3 billion in private investments and $27 billion in contingent commitments across model developers (OpenAI, Anthropic), cloud operators (CoreWeave, Nebius Group), and technology suppliers.
Why it matters
Specialized cloud operators that depend on external funding face risk if financing dries up. CoreWeave spent $6.8 billion on capital expenditures against $2.1 billion in revenue in Q1 2026; Nebius spent $2.5 billion against $399 million in revenue. Both have massive backlogs ($99.4 billion and $4.8 billion respectively), but the timing mismatch means they must fund infrastructure before customer cash arrives.
What to watch
Companies with large revenue backlogs but heavy near-term capital needs—CoreWeave, Nebius, and Iren (which Nvidia has not yet fully funded)—could face slower growth if external financing becomes harder to secure. Large, profitable players with existing cash flow are less vulnerable.
On July 28, Mark Cuban raised concerns about how the artificial intelligence boom is being financed, comparing Nvidia's role to the IPO machine of the dot-com era. Cuban was not claiming that Nvidia literally takes companies public; rather, he was drawing a parallel to the 1990s internet boom, when IPOs gave young internet companies the capital to expand rapidly. Today, Nvidia is making direct investments in AI model developers and cloud operators, effectively funding the sector's growth in much the same way.
Nvidia's investment footprint is substantial. As of the end of the first quarter of fiscal 2027 (April 26, 2026), the company held $42.3 billion in private investments and another $27 billion in contingent investment commitments. Its portfolio spans model developers including OpenAI and Anthropic, cloud operators such as CoreWeave and Nebius Group, and technology suppliers like Intel, Synopsys, Nokia, and Coherent. While some of these deals benefit Nvidia twice—the investment rises in value and the recipient may buy more Nvidia technology—the relationship is not automatic. Many investments support suppliers and partners rather than direct customers, so Nvidia's portfolio alone does not guarantee additional sales.
Cuban's warning is not a blanket argument against AI stocks. Large cloud providers and profitable AI chip and networking players can fund much of their spending from their existing operations. A slowdown in data center construction could dent their cash flow or valuations, but it would not immediately threaten their viability. The warning applies most urgently to specialized cloud operators that depend on regular outside funding to keep expanding. CoreWeave and Nebius exemplify the risk. In the first quarter of 2026, CoreWeave generated nearly $2.1 billion in revenue but spent $6.8 billion on capital expenditures. Nebius showed a similar imbalance, with $399 million in revenue against nearly $2.5 billion in capital expenditures in the same period. Demand is not the problem—CoreWeave exited the quarter with a $99.4 billion revenue backlog, and Nebius had nearly $4.8 billion in deferred revenue. The challenge is funding the GPUs and data centers needed to deliver that future revenue before the cash from customers arrives.
Iren, another close Nvidia partner, faces a comparable timing issue at an earlier stage. The company generated $144.8 million in revenue in the third quarter of fiscal 2026 (ending March 31, 2026) but spent about $1.36 billion on computer hardware, property, and equipment. Notably, Nvidia has not yet deployed its full agreed-upon $2.1 billion investment in Iren; instead, Nvidia holds the right to purchase up to 30 million Iren shares at $70 each, subject to certain conditions. This partial commitment underscores the broader point: Nvidia's stake in these companies is only part of the story. The larger issue is whether specialized cloud operators can eventually fund their expansion with cash from their own businesses. Companies that still depend on external financing could face slower growth if funding becomes harder to obtain.
Mark Cuban's comparison of Nvidia to a dot-com IPO machine captures a structural shift in how the AI sector is being financed. During the 1990s internet boom, IPOs provided young companies with capital to expand before they were profitable; today, Nvidia is effectively playing that role by investing across AI infrastructure and model development. With $42.3 billion in existing private investments and $27 billion in contingent commitments, Nvidia has become the sector's de facto funding engine.
The risk Cuban highlights is real but concentrated. Large, already-profitable players—mature cloud providers and established chip makers—can self-fund their expansion. They are insulated from sudden financing shocks because their own cash flow supports growth. The vulnerability lies instead with specialized operators that are growing faster than their revenue can support. CoreWeave and Nebius exemplify this gap: both are spending 2–6× their quarterly revenue on capital expenditures, betting that customer orders in their backlogs will eventually generate the cash to repay those upfront costs. That bet is rational if demand holds and financing remains available, but it becomes dangerous if either factor shifts. Iren, a smaller Nvidia partner not yet fully funded by Nvidia, faces the same timing problem at an earlier stage. For these companies, the issue is not demand—their revenue backlogs are substantial—but whether they can bridge the gap between infrastructure spending and customer payments without a steady stream of external capital.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
TSMC has roughly doubled its CoWoS and SoIC packaging capacity annually for several years and is now running C…
Tech companies are raising unprecedented sums for AI infrastructure—$194 billion so far in 2026 by just four f…
Broadcom CEO Hock Tan reaffirmed in early June that the company expects fiscal 2027 AI semiconductor revenue o…
Nvidia has partnered with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to raise $500bn in…
Nvidia has signed letters of intent with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to…
Anthropic has signed the EU AI Act Code of Practice and will embed invisible watermarks in Claude-generated te…
The AI news that matters, in one minute each morning.
Sign up free