By Wednesday night this week, you’ll know whether the AI build-out is still a bull market or just a math problem.
Microsoft, Alphabet, Meta, and Amazon all report after the close on Wednesday. Apple follows on Thursday. The Fed makes its rate decision the same Wednesday afternoon, with Jerome Powell’s last meeting before his term ends in May. Combined market cap on the move: roughly thirteen trillion dollars. Combined 2026 AI capital expenditure already announced or implied across those four hyperscalers: north of four hundred billion.
That number used to sound like science fiction. Two years ago, Meta’s full-year capex was about $30 billion. The midpoint of its 2026 guide is around $125 billion. Alphabet roughly doubled its 2025 capex into 2026. Microsoft is on pace for $90B+ this fiscal year, most of it AI infrastructure. Amazon’s AWS capex line has gone vertical. None of these companies is spending this kind of money to hedge their bets. They are spending it because they have decided AI is the platform shift that defines the next decade, and they would rather over-build than miss it.
The question the market is asking on Wednesday night is the question it has been asking, more loudly, every quarter since the start of 2026: when does the revenue catch up?
The capex-revenue gap is real, and it is getting bigger
Here is the part most people are not internalizing. The four big hyperscalers will spend somewhere around $400 billion on AI infrastructure in 2026. The entire enterprise software industry’s revenue from generative AI products in 2026 is, by most credible estimates, in the $50–80 billion range. Even if you generously include consumer ChatGPT-style subscriptions, you do not get close to $400 billion in AI revenue against $400 billion in AI capex.
That gap does not have to close in one quarter. AI infrastructure has a long depreciation schedule, and the assets being built today will produce revenue for years. But the gap has to be visibly closing each quarter, with a credible narrative for when the lines cross. That is why this week’s prints matter so much. If Azure’s AI-attributed growth accelerates, Copilot seat counts ramp, Meta’s AI-driven ad pricing climbs, and AWS gives its clearest picture yet of generative AI contribution, the bull case holds. If any one of those numbers stalls, the conversation changes overnight.
What is actually worth watching on Wednesday night
For Microsoft: Azure constant-currency growth, but split into AI-attributed versus non-AI. Copilot enterprise seat counts, if disclosed. Capex guide for FY27. Most readers will fixate on the headline EPS. The professionals will fixate on whether the AI chunk of Azure is still the fastest-growing line item.
For Alphabet: Search revenue resilience as AI Overviews continue to expand. Google Cloud margins. Capex revision. The bear case on Alphabet has always been that AI cannibalizes search; the bull case is that AI cannibalizes search but Google captures the new search anyway. Wednesday’s print is another data point.
For Meta: Ad pricing. AI is supposed to make Meta’s ad targeting better, not just cheaper. If average price per ad is climbing while volume holds, the AI-into-ads thesis is intact. Capex commentary is the swing factor.
For Amazon: AWS growth rate, AWS operating margin (this is the cash machine that funds everything else), and any color on Trainium adoption. AWS has been gaining share against Azure on AI-specific workloads at the inference end of the stack.
For Apple on Thursday: services growth, iPhone units, and any signal about the on-device AI strategy. Apple is the odd one out here because its AI thesis is consumer-facing and hardware-bundled, not enterprise capex.
The Fed is the side dish, not the main course
Polymarket has the Fed at near 100% no-cut for Wednesday. Fed funds futures imply policy stays on hold through the rest of 2026. The Cleveland Fed nowcast for headline inflation has actually been climbing in recent weeks, not falling. The market is in the awkward position of having priced two rate cuts this year while the underlying data argues for zero.
That mismatch is the press conference’s trap. Powell has every reason to lean hawkish into a strong labor market and a re-accelerating inflation print. If he does, every duration-sensitive trade in the market wobbles for forty-eight hours. Then earnings hit. The setup is built for volatility, both up and down.
For long-term wealth builders, this matters less than the headlines suggest. Two-year Fed paths rarely matter to compounding outcomes over twenty years. The hyperscaler capex cycle, on the other hand, will determine which mega-cap names are still leading the index in 2030. That is the trade that compounds.
The contrarian read
Most coverage of this week will be framed as risk: bubble talk, capex skepticism, the “what if AI demand is inflated” angle. Some of that is healthy. Anthropic’s leadership has been openly more conservative on AI demand than the OpenAI/NVIDIA narrative implies, and that is worth listening to.
But the contrarian read inside the contrarian read is this: even if AI demand is somewhat inflated relative to the most aggressive forecasts, the absolute level of demand is still enormous, and the supply side is still fully sold out for years. NVIDIA’s data center backlog, TSMC’s CoWoS-L allocation, the global power capacity bottleneck — none of that has loosened. The bull case does not require AI demand to be perfect. It just requires it to be roughly real, which the prints will either confirm or break.
What a wealth builder actually does this week
Three things, in order.
First, do not trade the print. Most retail money that “trades earnings” loses to the spread, the implied volatility crush, and the after-hours whipsaw. If you are a long-term holder of any of these names, your job this week is to read the calls and update your thesis, not to flip the stock.
Second, look at your AI exposure as a basket, not as single names. The AI build-out is going to produce winners across at least four layers — chips, infrastructure, hyperscalers, and applications. Concentration risk in any single name is the single biggest mistake retail makes in this cycle.
Third, write down what would change your mind. If Microsoft’s AI-attributed Azure growth decelerates two quarters in a row, what does that mean for your AI thesis? If Meta cuts capex guide, what does that signal for the rest of the cluster? Pre-committing to the data points that would update your view is the thing that separates wealth-building investors from investors who are, in practice, just gambling.
Bottom line
This week is the cleanest read on the AI capex cycle the market has had since the cycle started. The numbers are big enough to matter, the timing is tight enough to compress every variable into forty-eight hours, and the press conference adds a macro overlay that will distort the first reaction. By Friday, the market will know whether the AI bull case is still working, or whether 2026 starts to look more like a digestion year. Either outcome is investable, but only if you have decided ahead of time how you would respond.
Track your own progress against these macro setups instead of being reactive to them. The Wealtharian Wealth Tracker lets you monitor your net worth, FU money number, and investment milestones in one place — so when weeks like this one move the market, you know exactly how it moves you. Try it free →