The April 2026 tech earnings bonanza put the AI investment thesis under a microscope. Four of the world's largest technology companies — Alphabet, Amazon, Meta, and Microsoft — are on track to spend a combined $725 billion on AI infrastructure in 2026 alone. The question that dominated analyst calls and post-earnings commentary: is any of that spending actually creating tangible benefits?

The answer, based on the latest results, depends heavily on which company you ask.

Alphabet emerged as the clear winner. The stock hit a new record high after Google reported strength in its cloud division, with cloud revenue growing 60% year over year. The company also revealed a backlog of AI-related commitments that doubled quarter over quarter, according to Bloomberg's Caroline Hyde. Brad Aaronson, a portfolio manager quoted on the show, said the spending is "absolutely" justified because of "exponential demand" for AI through Google Cloud. He noted that Google has positioned itself as a low-cost token provider for AI, and the vertical integration of its business model — from chips to services — is creating a flywheel effect.

Amazon also delivered an acceleration in AWS growth, which Ed Ludlow described as "not seen in quite some time." The stock, however, came off near-record highs, possibly because investors already priced in the momentum. Amazon is another company that, like Alphabet, can point to a direct revenue line from AI: businesses renting cloud compute and AI tools.

Microsoft posted a "very solid earnings growth number," according to Aaronson, but the stock fell more than some expected. The issue wasn't the quarter itself. It was guidance. Microsoft forecast a "modest acceleration" into the second half of the year, which landed with a thud in a market looking for a slam dunk. The company also got caught in a broader software downdraft that has hit the sector in early 2026. Still, Aaronson called the weakness a buying opportunity, arguing that Copilot will be one of the first AI use cases enterprises adopt.

Meta was the outlier in the worst way. The stock was down 9% after hours, and down 10% in the following trading session. The company issued revenue guidance for the current quarter of up to $61 billion, essentially in line with consensus. But it did not provide a single metric showing that its massive AI investment is paying off. The company is on track to spend up to $145 billion in infrastructure this year — more than any of its rivals relative to revenue — yet it has no cloud business to sell compute through.

Mark Zuckerberg said on the call that most of the capex increase is driven by higher component costs, particularly memory pricing. He expressed confidence in the investment but emphasized a focus on efficiency. Investors didn't buy it.

Brad Aaronson explained the dynamic: "Each quarter that ticks off with all of these companies, fair or unfair, we have to evaluate what have you done for me lately relative to that spend." For Meta, the revenue was fine but decelerating in Q2 guidance. And while Alphabet and Amazon saw their capex guidance rise partly due to acquisition or demand, Meta's increase was purely organic and cost-driven.

Ed Ludlow raised a critical point: "Meta doesn't have the sales channel." Alphabet can point to Gemini's 40% growth in enterprise use; Amazon and Microsoft can point to cloud contracts. Meta's AI is free inside Instagram and Ray-Ban smart glasses. Zuckerberg can argue the technology improves ad targeting and content recommendations, but that's a hard sell when you're spending $145 billion.

"Clearly the commercial cloud vendors are the ones showing the best return, particularly with that margin upside," Aaronson added. Meta, by contrast, is in a tougher spot: decelerating revenue, rising costs, and no direct AI revenue line.

The capital expenditure story has a second-order effect. If all four companies are raising their spending forecasts, one would expect NVIDIA, the dominant supplier of AI chips, to be the biggest beneficiary. Yet NVIDIA shares fell. Why? Because the hyperscalers are increasingly talking about building their own custom silicon.

Aaronson noted that the shift toward inference — running already-trained models — benefits companies like Broadcom, which designs custom chips. "Perhaps people are thinking that NVIDIA is not going to be the beneficiary as this infrastructure continues to build out," he said. NVIDIA will still be essential for training new models, but the inference market is opening up to alternative architectures including CPUs.

On the same program, Bloomberg reported that Anthropic has begun weighing a fresh funding round that could value the company at $900 billion or more, potentially overtaking OpenAI. No term sheet has been signed. The valuation is even higher than the $900 billion figure previously discussed. One reporter described it as "the standard dance in AI fundraising these days" — investors approaching Anthropic with capital offers before it goes public, and Anthropic responding to those offers rather than actively seeking them. The company has already taken strategic investments from Google and Amazon.

In a separate development, the NSA has been testing an Anthropic model to find vulnerabilities in its own software. According to a U.S. official, the agency is impressed by the model's speed and efficiency in searching for potential flaws. The White House is also preparing a national security memo that would set a standard for how federal agencies can work with AI companies. The memo aims to address tensions between the Pentagon and Anthropic, including measures that would force agencies to diversify the model companies they work with — effectively providing a workaround for the Pentagon to avoid fully exiting Anthropic while still managing supply chain risk.

Elon Musk returned to the witness stand in his lawsuit against OpenAI, becoming visibly irritated during cross-examination about the exact amount he contributed to the startup. His testimony is expected to wrap up, followed by his longtime fixer.

Stripe announced a partnership with Google at its annual conference, allowing businesses to use Gemini AI for commercial transactions. Stripe president and cofounder John Collison described the vision as starting with small, low-risk decisions — like buying a domain name — and working up the trust curve. "That has never been the case in technology when you get a truly more convenient way of working for consumers, they take the less convenient way. That has never happened," he said.

Caroline Hyde summarized the big picture: the four largest tech companies are constrained by capacity. They cannot build infrastructure fast enough to service external demand. At the same time, component prices are rising — memory chips, energy, and other inputs are all getting more expensive. "Less bang for the buck" when memory prices have soared. All of this squeezes free cash flow, even as revenue grows double digits.

Ed Ludlow identified free cash flow as the biggest issue for all four companies. "As the spend continues to rise despite these double-digit revenue growth numbers they are posting, the free cash flow continues to dwindle." The question investors must answer: how much are you willing to pay for these companies while cash flow remains under pressure, and how long will that pressure last?

For now, the market is rewarding companies that can point to a direct AI revenue line. Alphabet and Amazon have it. Microsoft is promising acceleration. Meta does not have it yet, and that impatience is showing up in the stock price.