Goldman Sachs Doubles Down on Silicon
The market expected a retreat. The bears predicted a cooling of the silicon fever. They were wrong. Goldman Sachs released a note on May 2, 2026, confirming that the AI trade remains not just intact, but structurally reinforced. Anshul Sehgal, global co-head of Fixed Income, Currency and Commodities at the firm, pointed to the latest megacap tech earnings as the definitive proof. The numbers do not lie. The capital expenditure is accelerating. The revenue streams are finally catching up to the hype. This is no longer a speculative bubble. It is an industrial overhaul.
Big Tech is trapped in a cycle of mandatory spending. To stop now is to surrender the future. Microsoft, Alphabet, and Meta have signaled that their commitment to AI infrastructure is absolute. This commitment is reflected in the latest Q1 2026 filings, where capital expenditures (CAPEX) have reached levels previously reserved for national defense budgets. The market is ignoring the skeptics because the cash flow remains robust enough to fuel the fire.
The FICC Perspective on Megacap Tech
Why is a Fixed Income and Commodities head talking about tech stocks? The answer lies in the plumbing of the global economy. AI is no longer a software story. It is a commodity story. It is about power, copper, and debt. The massive scale of data center construction has shifted the AI trade into the realm of macroeconomics. Goldman’s focus on this sector through the lens of FICC suggests that the volatility of these tech giants now impacts currency and bond markets as much as traditional sectors.
The cost of capital is the new bottleneck. While interest rates have stabilized in early 2026, the sheer volume of debt required to fund these GPU clusters is staggering. Investors are no longer looking at user growth. They are looking at the efficiency of the burn. They are looking at the yield on every dollar of silicon. According to recent Bloomberg market data, the correlation between tech earnings and energy futures has reached an all-time high. If you want to track the AI trade, you don’t look at the App Store. You look at the power grid.
Quarterly Financial Performance of AI Leaders
The following table outlines the reported CAPEX and Net Income for the top three infrastructure spenders as of their May 2026 disclosures. The scale of investment is unprecedented in the history of the private sector.
| Company | Q1 2026 CAPEX (USD Billions) | Net Income (USD Billions) | AI Revenue Contribution (%) |
|---|---|---|---|
| Microsoft | 17.4 | 24.2 | 31% |
| Alphabet | 14.2 | 21.8 | 19% |
| Meta | 12.5 | 15.4 | 24% |
The data suggests a narrowing gap between investment and return. In 2024, the AI revenue contribution was a rounding error. Today, it is a core pillar of the balance sheet. This shift justifies the aggressive stance taken by Goldman Sachs. The trade is intact because the utility is real. The software layer is finally monetizing the massive hardware investment of the previous two years.
Visualizing the Infrastructure Surge
The chart below illustrates the shift in capital allocation among the tech elite. We are seeing a pivot from stock buybacks toward hard infrastructure. This is a fundamental change in how these companies manage their cash piles.
Megacap AI Infrastructure Spending Q1 2026
The Energy Bottleneck and Margin Pressures
The trade is intact, but it is not without risk. The primary threat is no longer a lack of demand. It is a lack of supply. Not of chips, but of electrons. The energy requirements for the next generation of LLMs are exceeding the capacity of local grids in Northern Virginia and Dublin. This is where the FICC expertise becomes critical. We are seeing tech companies enter the long-term energy derivative market to hedge against rising electricity costs.
Margins are being squeezed by the operational costs of these data centers. It is one thing to buy the GPUs. It is another thing to keep them cool. The 10-Q filings from this quarter show a significant uptick in utility expenses. If the AI trade is to survive the rest of the year, these companies must prove they can optimize the inference costs. The training phase was expensive. The inference phase is where the profit is won or lost.
The Next Milestone
The market is now fixated on the upcoming Nvidia H300 shipment data due in mid-May. This will be the ultimate test of the infrastructure appetite. If the orders remain backlogged through 2027, the Goldman Sachs thesis will be vindicated. If we see the first sign of a cancellation, the FICC desks will be the first to sell. Watch the 10-year Treasury yield on May 15. If it spikes alongside tech volatility, the AI trade is no longer a growth play. It is a macro hedge. The era of easy software margins is over. The era of the industrial AI machine has begun.
