There is something about the numbers coming out of Silicon Valley in 2026 that is nearly impossible to understand. NVIDIA reported quarterly revenue of $81.61 billion, an 85% increase from the previous year. Azure from Microsoft increased by 40% in just one quarter. Alphabet increased its full-year capital expenditure guidance to between $180 and $190 billion after expanding Google Cloud by 63%. Almost as an afterthought, the company raised $31.1 billion in senior unsecured notes to help cover the cost. These are not typical business figures. They fall into an entirely different category of economic events.
However, a bond market that is flashing signals that serious investors find difficult to ignore sits just behind all of this, like a storm visible on the horizon that half the room refuses to look at. On May 21, the 30-year Treasury closed at 5.11%, right at the top of its trailing 12-month range. The yield over a ten-year period increased to 4.58%. The ratio of global debt to GDP is now 310%. These two realities—the surge in AI infrastructure and the sovereign debt pressure that is subtly growing beneath it—have led to something truly peculiar in the financial markets: intelligent, knowledgeable individuals examining identical data and coming to nearly incompatible conclusions.
Investor David Friedberg sat opposite Gavin Baker and Chamath Palihapitiya on a recent episode of the All-In Podcast, where the three of them essentially described three different versions of the same world. This tension was demonstrated in real time. Friedberg has the most pessimistic view. He thinks the spending issue will eventually break, not only in Washington but at every level of government in almost every nation. He is observing the long end of the Treasury curve and interpreting it as a market that is reluctantly and slowly starting to price in a real reckoning. Somewhere on his list of catalysts is a possible unwind of the Japanese carry trade, a slow-moving risk that could abruptly accelerate in ways that few portfolios are designed to handle.
Baker makes a strong pushback. He contends that America’s dominance in AI infrastructure and energy independence amount to something historically significant, not just another tech cycle but a structural shift in which nation controls the most important economic levers of the coming decades. It’s interesting to consider his comparison to Cisco during the dot-com bubble: Cisco once traded at 100 times forward earnings due to speculation. In contrast, NVIDIA trades on actual, verifiable cash flows at a forward multiple in the mid-teens. According to Baker, the AI case has a stronger foundation than the tech case from the late 1990s.
Whether Baker’s optimism adequately explains what happens to capital expenditure plans in the event that the 30-year Treasury moves significantly above 5.5% is still up for debate. In just one quarter, Microsoft made capital expenditures totaling $30.88 billion. For the entire year, Alphabet is aiming for close to $190 billion. When money was cheap, these figures made sense. Even though share prices indicate it’s leaning toward yes, the market hasn’t fully responded to the question of whether they still make sense with long rates sitting near two-decade highs. Microsoft’s 13% year-to-date decline indicates that at least some investors are accounting for their anxiety while still holding the position.
Perhaps sensibly, Palihapitiya declines to properly plant his flag in either camp. Recognizing the macro stress while remaining committed to the core winners of the AI buildout, he focuses on five or fewer high-conviction names. That position has a certain intellectual honesty to it, acknowledging that both can be true at the same time, that rates can be dangerously high, and that AI fundamentals can still be truly exceptional. There is no neat resolution to that tension. It simply needs to be controlled.
It’s difficult to avoid thinking that the macroeconomic divide of 2026 isn’t primarily about technology versus finance as you watch all of this happen. It has to do with timing. There are actual AI factories. There is a real risk of a debt spiral. The question is which comes first: the bond market moment that compels everyone to halt and recalculate, or the productivity gains that justify the spending. This is the biggest infrastructure expansion in human history, according to Jensen Huang. He might be correct. Whether the world can afford to finish it will be determined in part by the yield on the 30-year Treasury.
