Google Cloud faces 48% revenue reliance on OpenAI and Anthropic compute partnerships in 2027, creating single-customer risk.

AI EZ Primary Research chief executive Ed Zitron used a Bloomberg interview published on July 31, 2026 to argue that the capital expenditure wave lifting technology stocks rests on two loss-making customers. Citing UBS estimates, Zitron contended that OpenAI and Anthropic will account for 27% of Google Cloud revenue this year and more than 48% next year. The interview, released on Bloomberg Podcasts' YouTube channel following quarterly earnings reports from Microsoft, Amazon, Meta, and Alphabet, drew more than 420,000 views by August 2. "Everyone is buying into these stocks because they believe all of that CapEx is going towards diverse and spread out AI demand, when in fact, what it's actually doing is helping create infrastructure for two unprofitable, unsustainable companies," Zitron said.

The figures underlying the claim are rarely disclosed. According to UBS estimates cited by Zitron, OpenAI and Anthropic together will generate more than $124 billion in Google Cloud revenue next year, with Anthropic alone contributing $76 billion in 2027. Zitron noted that OpenAI's position as a large Google Cloud customer remains little known, a detail he attributed to UBS analyst Stephen Ju.

Microsoft follows a similar pattern. Zitron cited Barclays estimates placing the two AI companies at 13% of cloud revenue this year and 18% next year for what he described as a much larger business than Google Cloud. According to his own reporting on OpenAI's finances, 69% of year-over-year growth in the Microsoft Intelligent Cloud segment during 2025 came from OpenAI. Without that single customer, the segment would have grown 8%, barely beating inflation.

The revenue concentration extends beyond the hyperscalers. According to The Information, 89% of revenue at the largest AI companies comes from OpenAI and Anthropic alone. When asked whether such concentration is necessary given the cost of building capable models, Zitron clarified that his concern targets revenue concentration in two companies rather than compute itself.

The scale of projected payments raises obvious questions. "How is Anthropic going to afford that? They burned tens of billions of dollars," Zitron said, referring to the $76 billion UBS estimate for 2027. He brought direct knowledge of OpenAI's accounts to the discussion, having reported the company's audited financials for the Financial Times. "It's a company just burning cash. They lost $20.9 billion in 2025," he said. More than $800 million of OpenAI's revenue that year came from SoftBank for a program called Crystal Intelligence, according to Zitron, who said he could find no evidence of activity connected to it. SoftBank holds a large shareholding without board seats.

OpenAI's delayed listing compounds the pressure. The New York Times reported the company is considering postponing its public offering until 2027. "That's lethal for a number of people," Zitron said. The structural problem sits in the cash cycle: "OpenAI and Anthropic need continual flows of capital. They do not pay their bills out of existing cash flow. Any interruption to that capital becomes the first domino."

On due diligence, Zitron offered a different reading than negligence. The hyperscalers "did the due diligence in the sense that they said, we are going to create our largest customers and we're going to own large parts of them," he said. The arrangement extends to hardware. According to Zitron, Broadcom sells tensor processing units to Google, which then sells them to Anthropic and rents the chips back through Google. "Google gets to double up on revenue," he said.

The funding picture follows the same loop. OpenAI and Anthropic have raised between $200 billion and $300 billion, yet effective totals run higher because Microsoft, Google, and Amazon built the underlying infrastructure. Testimony from the trial between Sam Altman and Elon Musk revealed a Microsoft executive placing infrastructure cost at $100 billion, settling on roughly $70 billion to $80 billion of capacity the AI companies never paid for.

The customer concentration risk has already reached credit markets. S&P Global Ratings cut Oracle to BBB- in July 2026, noting that OpenAI accounts for roughly half of the $638 billion in remaining performance obligations Oracle carries.

On the demand side, Zitron worked through the arithmetic. Sightline Climate identified about 190 gigawatts of data center capacity in planning or construction as of February. Applying a power usage effectiveness rating of 1.3 and $12 million per megawatt, he calculated that facilities would require more than $1.6 trillion in annual revenue to justify themselves. "Having two customers is not going to do that," he said. Even OpenAI's spending would not suffice; he estimated the company would need $400 billion annually, funding he doubts it will secure.

Construction timelines add friction. Data centers take 12 to 36 months to build depending on size, slower than market narratives assume.

The transformation of the platforms themselves troubles Zitron equally. Amazon, Google, and Microsoft have shifted from cash-generating businesses with low asset intensity into what he called "bulbous, GPU filled asset mongers," filled with semi-built data centers serving two or three customers. Meta invests in AI but does not yet sell compute capacity. For Zitron's thesis to hold, nothing dramatic needs to occur; OpenAI and Anthropic would simply need to grow to implausible size to make capacity pay off, because otherwise demand for compute at scale does not exist.

The companies in question sell most of the world's advertising, and the spending Zitron questions runs through earnings reports marketers watch quarterly. Alphabet raised its 2026 capital expenditure guidance to $195 billion to $205 billion on July 22, 2026, reported negative free cash flow of $5.9 billion for the quarter, and carries $98.2 billion in long-term debt, up from roughly $16 billion a year earlier. The company raised approximately $85 billion in equity in June 2026 to fund infrastructure described as supply-constrained against demand. Meta reported quarterly capital expenditures of $31.08 billion on July 30, 2026, against $17.01 billion a year prior, with free cash flow falling to $784 million. Microsoft's capital expenditures rose 70% to $41.0 billion in its June quarter, with calendar 2026 spending expected around $175 billion. Advertising revenue now funds infrastructure whose demand case rests, in Zitron's account, on two unprofitable tenants.

The exposure runs bidirectionally. The same AWS infrastructure serving Claude and GPT-5 powers programmatic bidding, creative generation, and Amazon's Rufus shopping assistant. Operational continuity of AI ad tooling depends on economics Zitron disputes. OpenAI itself has entered the advertising market that would need to fund it, with projections of $102 billion in advertising revenue aimed at competing with Google's $224 billion search business.