Is the AI Party Nearing Its First Reality Check as Costs Rise and Spending Questions Grow?

Three years of AI spending. Record stock prices. Infrastructure buildout running at full speed. And not one of the biggest companies in the trade can draw a straight line between what they are paying and what they are getting back. That is not my opinion. That is Uber‘s chief operating officer Andrew Macdonald saying it out loud on a podcast this week. The spending is easy to count. The returns are the part nobody has figured out yet.

 

Ahmed Yousre, Global Market Strategist at PU Prime commented:

The artificial intelligence (AI) theme continues to be one of the strongest drivers supporting U.S. equities, particularly within the technology and semiconductor sectors. Strong capital expenditure commitments from major technology companies have fueled significant investment into data centers, cloud infrastructure, advanced semiconductors, and enterprise AI solutions, helping sustain bullish sentiment across Wall Street.

However, from our perspective, the market may be approaching a new phase of the AI cycle. While investor enthusiasm has largely been driven by spending growth and infrastructure expansion over the past several years, attention is gradually shifting toward the ability of companies to generate measurable returns from these investments.

Recent discussions surrounding AI spending efficiency suggest that some corporations are becoming increasingly focused on cost management and return on investment rather than pure adoption metrics. This does not necessarily signal a slowdown in AI development, but it may indicate that future market performance will depend more heavily on earnings delivery, productivity improvements, and commercial monetisation.

At the same time, investors should not overlook broader macroeconomic risks. Recent U.S. economic data has remained relatively resilient, supporting Treasury yields and reducing expectations for aggressive Federal Reserve rate cuts. A higher-for-longer interest rate environment could continue creating valuation headwinds for high-growth technology companies, particularly those trading at elevated multiples.
While the long-term AI growth story remains compelling, market participants may become increasingly selective moving forward.

Companies capable of translating AI investment into sustainable revenue growth and profitability are likely to continue attracting institutional capital, while firms struggling to demonstrate tangible returns may face increased scrutiny.

Overall, we believe the AI theme remains supportive for equities, but investors should closely monitor both earnings performance and macroeconomic developments, as rising yields and valuation concerns could gradually limit the pace of future gains.

$700 Billion in AI Spending and the Biggest Buyers Want Out of the Pricing

Nvidia remains above its rising long-term trend, but the stock has struggled to build on strong earnings and guidance, suggesting investors may be becoming more selective toward AI leaders. Source: TradingView.

AI infrastructure spending is tracking toward $600 billion to $700 billion in 2026. Data centers. Chips. Computing capacity. That number carried the bull case for three years straight. More spending. More demand. More room to run. Nvidia rode that wave harder than anyone.

But inventory levels at Nvidia and AMD are running above normal ranges in recent analyst discussions. Some customers may need to digest what they already purchased before placing new orders. That is not demand falling apart. That is a normal pause after three years of aggressive buying.

The signal that matters more is coming from the customer side. Google, Amazon, and Microsoft are all building custom AI chips in-house. All three of them at the same time. They did not decide to become chipmakers because they thought it sounded interesting. They looked at what they are paying Nvidia and concluded they would rather invest the engineering money than keep writing checks at that price. The companies buying the most chips in the world are actively working to stop buying them at the current cost. The order books may still look full today. The pricing relationship is already shifting underneath.

Microsoft Cut Claude Code Licenses and Moved to Cheaper Tools

Microsoft hit the same wall. The company expanded Claude Code access to developers in December 2025. Usage climbed fast. By mid-2026 Microsoft had reportedly pulled most of its direct Claude Code licenses in key departments and moved developers over to GitHub Copilot CLI. Their own tool. Lower cost.

Think about what that means. Microsoft spends tens of billions on AI infrastructure every single quarter. Deepest pockets in the industry and it is not even close. If that company is cutting back on a coding tool because the token costs got uncomfortable at scale, the economics are telling you something. AI demand is not falling off. But the companies with the most money to spend are starting to watch the operational costs, not just the infrastructure costs. That is a different conversation than the one the market has been having for three years.

Uber’s COO Says the ROI Link Is Not There Yet

Uber handed roughly 5,000 engineers access to Claude Code late last year. They did not have to push hard to get adoption. The company ran an internal leaderboard ranking teams by AI tool usage. Engineers started competing on it. Code writing. Debugging. Development work across the board. Usage went through the roof. By early 2026 Uber had burned through its entire annual AI coding tools budget in four months flat. Four months. Some of the heavier users were generating costs between $500 and $2,000 per month on tokens alone. Multiply that across thousands of engineers and the number gets uncomfortable fast.

Then Macdonald went on the Rapid Response podcast and said the thing that should matter to anyone holding AI exposure. He said it is very hard to draw a line between the usage numbers going up and actual useful consumer features getting shipped. Not a budget complaint. A senior executive at one of the most AI-forward companies in Silicon Valley telling investors the output is not matching the input. The tools get used constantly. Engineers are moving faster. But when Uber’s leadership tries to find the consumer-facing improvement that justifies $951 million in first-quarter research and development spending, up 17% from the year before, they come up empty. The money is going out the door. What is walking back in is harder to see.

The IPO Wave Could Become the Next Test

OpenAI is reportedly targeting a public offering as early as September. Anthropic is widely expected to follow later in the year. Private capital keeps flowing in at enormous valuations. Public investors are a different audience.

For most of the AI boom the market rewarded growth and did not ask many questions about returns. User numbers went up. Data centers came online. Stocks climbed. Now Uber says the link between AI spending and consumer results is not clear. Microsoft retreated from one of the most expensive tools in the stack. Both of those stories landed in the same month. The IPOs could arrive right when investors start wanting to see profitability instead of promises.

These could be some of the largest technology offerings in years. Investors will want in. The question is where the money comes from. Institutions are not sitting on unlimited cash. Some of it arrives as new capital. Some of it gets pulled out of positions already sitting in AI portfolios. The risk is not that the deals fail. The risk is that investors fund them by selling something else in the AI complex to free up the cash. Strong demand says the market still wants to pay for expansion. A weak reception says the patience ran out.

Technical Picture Still Bullish, But Less Uniform

AMD keeps pushing higher near record territory. Buyers step in on every dip. The stock sits well above its long-term trend. That chart does not look like a trade that is breaking down. Source: TradingView.

Microsoft’s more muted price action contrasts with the strength seen in leading AI infrastructure names, reflecting a market that is becoming increasingly selective. Source: TradingView.

Microsoft is a different picture. Still trending higher but spending months going sideways while other AI names kept making highs. Nobody abandoned Microsoft. They just stopped chasing it with the same conviction. Nvidia has not been able to build on strong earnings and strong guidance. Three stocks. Three charts. Three different stories inside the same trade. The market is not walking away from AI. It is deciding which names deserve the money and which ones need to prove something first.

The Spending Story Is Shifting Toward a Returns Story

The AI trade is not broken. Adoption keeps expanding. Infrastructure demand holds up. The technology gets better every quarter. Nobody is leaving. But three years of rewarding companies for spending just because they were spending is starting to shift. Uber’s COO told investors the link is not there yet. Microsoft pulled back when the costs ran ahead of the value. Nvidia’s biggest customers are building their own chips to get off the current pricing.

None of that ends the story. It changes what investors watch next. Less focus on how much is being spent. More focus on what it is producing. The question stopped being whether AI is real a long time ago. The question now is whether the companies writing the biggest checks can show measurable returns before the market decides the promise alone is not enough anymore. Three years bought the buildout. The next three have to justify it.