OpenAI is reportedly burning $12.3 billion per quarter on infrastructure, while its created rival has just surpassed its own valuation metrics.
CNBC’s Kate Rooney recently walked viewers through OpenAI’s second-quarter financials, drawing on Wall Street Journal reporting for a company that does not file with the SEC and lacks public audits. The resulting picture is striking: OpenAI’s losses widened significantly while its revenue grew, yet rival Anthropic—founded by former OpenAI employees—quietly surpassed it on key metrics ahead of a potential public offering.
“OpenAI reportedly [is] seeing deeper losses in the second quarter, at least while sales grew sequentially,” Rooney noted. “That double digit growth was outshined by rival Anthropic, which is also poised to go public.” Because neither company trades publicly, ordinary investors cannot buy the winner or short the loser. Instead, the divergence serves as a signal for where AI infrastructure spending is heading, as those expenditures ultimately flow to the income statements of semiconductor firms supplying the underlying hardware.
OpenAI’s operating loss expanded to $12.3 billion for the quarter, rising from $9 billion in the prior period, inclusive of stock-based compensation. Revenue reached just under $7 billion, marking an 18% sequential increase. Growth accelerated in July to 20%, with enterprise revenue surging 32%.
In contrast, Anthropic achieved EBITDA profitability while generating $11.5 billion in quarterly revenue. Its valuation recently climbed to $65 billion, compared with approximately $40 billion for OpenAI—a clear inversion from their earlier market positions.
“The profit profile looks so different for these companies,” Rooney observed. “When you compare Anthropic, which at least on an EBITDA basis is profitable, to OpenAI, which has seen deeper and deeper losses, it calls into question some of the spending.” Investors should note that both are private entities; none of these figures represent audited public filings, and all should be treated as reported rather than confirmed.
OpenAI CFO Sarah Friar has publicly argued that compute capacity is the company’s primary moat. The thesis holds that more data centers yield greater capacity, which drives revenue and funds subsequent expansion. Rooney summarized the strategy: “OpenAI’s moat, as they’ve described it, is that they’re spending more on data centers, they want more compute, and that translates directly to revenue, which CFO Sarah Friar has laid out.” While the argument carries weight, Anthropic’s results provide the first serious evidence that massive scale may not be a prerequisite for success. If a competitor with a fraction of the infrastructure footprint can achieve EBITDA profitability at a comparable revenue base, the moat theory begins to appear speculative.
The comparison warrants careful interpretation. EBITDA excludes depreciation and stock-based compensation—costs that heavily dominate compute-intensive businesses—so labeling one company profitable and the other a loss-maker is not strictly an apples-to-apples assessment. Nevertheless, the trend is significant, and public-market investors historically reward narrowing losses more harshly than they penalize absolute loss size.
This AI capital cycle is primarily financed by private companies, yet the spending lands squarely on the income statements of a concentrated group of suppliers. Most visible are NVIDIA (NASDAQ:NVDA | NVDA Price Prediction) and Broadcom (NASDAQ:AVGO), though the buildout extends well beyond chipmakers into power, cooling, and networking sectors. NVIDIA reported $75 billion in data center revenue for its most recent quarter, a 92% year-over-year increase. Management identified both OpenAI and Anthropic as customers scaling operations on Grace Blackwell and Vera Rubin systems. Jensen Huang told analysts NVIDIA’s coverage of Anthropic “has been largely zero until just recently. And so we’re gaining share tremendously fast in inference.”
Broadcom disclosed a contractual obligation to deploy 1.3 gigawatts for OpenAI in 2027, part of a broader 10-gigawatt agreement extending through 2029. The company also secured access to over 1 gigawatt of TPU-based compute for Anthropic in 2026. According to CEO Hock Tan, AI semiconductor bookings exceeded $30 billion in the quarter against $10.8 billion in shipments.
Rooney highlighted a risk investors should not overlook: “If you do start to see more pushback, which it seems like there’s been this inflection this week, there’s been more of a grassroots effort to push back on data centers, this really could be a bottleneck for both of these companies. OpenAI [is] much more exposed here.”
Whether OpenAI’s widening losses reflect a deliberate strategic choice or a structural problem, suppliers continue to book revenue as orders persist. If the losses are structural, however, the entire supply chain will eventually absorb the impact, starting with firms whose backlogs assume current deployment rates will hold. At current trading levels, NVIDIA stands at $217.56 and Broadcom at $362.48, with both valuations embedding market assumptions about which scenario ultimately prevails.