AI Capex Is a Tax. The Chip War Just Leveled Up. And Powell Just Gave His Last Press Conference.
A quick note from Behind the Markets
Wall Street loves a simple story. "AI is the future." "Big Tech can't lose." "Just buy the leaders."
But when the leaders start spending like drunken sailors… that's not a victory lap. It's a stress test. For margins. For power grids. For supply chains. For the entire market's risk appetite.
And yesterday added a layer most investors haven't processed: Jerome Powell just held what is likely his final press conference as Fed chair. His term ends May 15. The Senate Banking Committee advanced Kevin Warsh's nomination to replace him on the same day. The era of Powell is ending — and the era of Warsh begins in a world of $107 oil, sticky inflation, and an AI capex arms race that's consuming 90% of Big Tech's operating cash flow.
1) Big Tech Earnings: AI Capex Is Starting to Look Like a Tax
This morning's setup is obvious: Amazon, Meta, Alphabet, and Microsoft all reported last night. That's the whole tape right now.
The trick is to stop thinking about "beats" and "misses." The real question is whether AI spending is still being treated as a growth engine… or if it's starting to be priced like an operating cost you can't control.
Reuters flagged re-emerging skepticism around Big Tech's "lavish" AI spend, with a Wall Street Journal report in the background about OpenAI missing internal targets for weekly users and revenue.
If the demand curve isn't ramping as fast as the compute bill, the market's next phase is simple: cloud margins compress, capital intensity rises, and the "safe" mega-caps trade more like cyclical industrials. These four companies have committed a combined $635–$700 billion in 2026 capex. Amazon is projected to post negative free cash flow of $17–$28 billion. Alphabet quadrupled its long-term debt in 2025 to $46.5 billion. And they're spending roughly 90% of operating cash flow on capex — leaving almost nothing for the buybacks and dividends that used to justify the premium.
One company that collects the AI infrastructure toll without the capex risk:
Company: Arista Networks (SYM: ANET)
The leading provider of high-speed data center networking switches and software — the plumbing that connects every GPU cluster in every hyperscaler data center.
Arista is currently trading around $164.40. When hyperscalers spend $700 billion on AI infrastructure, every data center needs networking. Arista's 400G and 800G switches are the standard in hyperscale deployments — Meta, Microsoft, and others are all major customers. The difference between Arista and the hyperscalers: Arista doesn't spend $200 billion on capex. It sells the networking equipment, collects high margins, and generates strong free cash flow. Revenue has been growing north of 20% year-over-year, and the company carries minimal debt. In a world where AI capex is becoming a tax on the spenders, Arista is the vendor that gets paid regardless of who wins the AI branding contest.
Bottom line: Don't just listen for AI buzzwords. Listen for discipline. When the spending turns into a "must-do" line item instead of a high-ROI bet, the whole leadership story changes.
2) The Chip War Just Leveled Up — And the Second-Order Trade Is the One You Want
While everyone argues about which AI model is winning this month, Washington is tightening the screws on the physical bottlenecks.
The MATCH Act — Multilateral Alignment of Technology Controls on Hardware — cleared the House Foreign Affairs Committee on April 22 in what lawmakers described as the "largest significant export control markup in the history of Congress." Twenty-plus export control bills advanced in a single session.
The MATCH Act would ban all exports of DUV immersion lithography systems to China — the "workhorses" of semiconductor manufacturing that China cannot produce domestically. It names five Chinese companies in statute: SMIC, Huawei, Hua Hong, CXMT, and YMTC, including all subsidiaries and affiliates. And it gives the Netherlands and Japan 150 days to adopt equivalent controls — or the U.S. expands the Foreign Direct Product Rule unilaterally.
Here's the detail that matters most for investors: the bill also bans servicing of existing machines in Chinese fabs. Maintenance is destiny. A $200 million lithography machine isn't set-and-forget — it requires continuous calibration, software updates, and field support from the manufacturer to maintain yield. Cut off servicing, and output quality degrades over time. That's not a headline event. It's a slow strangulation of capacity.
The revenue impact is real. Applied Materials, Lam Research, and KLA booked a combined $19 billion in China revenue in 2025. ASML derived 27% of its revenue from China in 2025. Lam Research reported China at 43% of Q1 fiscal 2026 revenue — $2.28 billion. These companies face near-term pain as China revenue drops. But the longer-term effect is structural: when global chip equipment flows get weaponized, the AI capex boom becomes more expensive and less predictable — and the end buyers who must secure supply (defense, critical infrastructure, aerospace) pay up and sign longer contracts.
One company positioned on the right side of the chip equipment chokepoint:
Company: Lam Research (SYM: LRCX)
A leading supplier of semiconductor manufacturing equipment — etch, deposition, and clean systems — with exposure to both the AI buildout and the defense/industrial electronics surge.
Lam Research is currently trading around $248.32. Yes, the company faces China revenue headwinds — management expects China to fall below 30% of revenue this year. But the offset is structural: every new TSMC fab in Arizona, every Intel foundry expansion, every Samsung facility in Texas needs Lam's equipment. The CHIPS Act is funding domestic capacity buildouts worth tens of billions. And the MATCH Act, by restricting China's access to equipment, accelerates the reshoring of advanced manufacturing to the U.S. and allied nations — all of which buy from Lam. The near-term China revenue hit is priced in. The multi-year domestic and allied buildout is not.
Bottom line: The chip war isn't a one-day headline. It's a slow tightening of the vise — and the winners are companies positioned around security-of-supply, not just the poster-child chip designers.
3) The Fed "Pause" Narrative Is Too Cute — And Powell Just Handed the Keys to Warsh
Yesterday's FOMC delivered exactly what the market expected: no change at 3.50–3.75%. But the subtext was anything but routine.
This was almost certainly Powell's final meeting as chair. His term expires May 15, 2026. The Senate Banking Committee advanced Kevin Warsh's nomination to replace him — splitting along party lines — on the same morning. Powell could stay on as a governor, but the leadership transition is now functionally in motion.
Here's what matters for independent investors: a Fed leadership change during a war, an oil shock, and an AI capex bubble isn't a footnote. It's a regime change in monetary policy. Warsh has been publicly more hawkish on inflation and more skeptical of the Fed's balance sheet expansion than Powell. If Warsh inherits a 3.50–3.75% rate, $107 oil, and core PCE at 2.7% (with Morningstar forecasting 3.6% for the full year), the market's assumption that the next move is a cut may need to be completely repriced.
Meanwhile, the real tightening is happening elsewhere: credit terms, risk budgets, and the willingness to finance long-duration dreams. Brent traded above $111 intraday on Tuesday. The conflict is in its ninth week. Production shut-ins averaged 9.1 million barrels per day in April. And the IEA has warned of an "unprecedented supply shock."
One ETF that owns the companies thriving in a "policy constraint" environment:
ETF: Energy Select Sector SPDR Fund (SYM: XLE)
The benchmark U.S. energy sector ETF — the asset class that outperforms when inflation is sticky, the Fed is boxed, and oil stays elevated.
XLE surged 38% in Q1 2026 while the S&P 500 fell. If the Powell-to-Warsh transition signals tighter policy, energy producers with low breakevens and high cash flow become even more attractive relative to growth stocks that need rate cuts. XLE is the instrument that benefits from the constraint — the one sector where "higher for longer" is a feature, not a bug.
Bottom line: Don't let a "no move" meeting hypnotize you. The constraint is policy + energy + supply chains — and now a new Fed chair who may be less inclined to rescue the market.
4) What's Worth Watching Into the Open: Profit-Taking Isn't a Signal — It's a Regime Change
Reuters quoted an analyst saying the OpenAI news created "convenient cover" to liquidate high-performing stocks and secure profits.
That's not just trader talk. It's the tell.
In a momentum regime, investors buy first and rationalize later. In a risk-off regime, investors sell first and demand proof later. The shift between those two regimes doesn't happen because of one headline. It happens when the market's tolerance for "faith-based investing" runs out.
And the evidence is accumulating. OpenAI missing targets. Amazon going negative free cash flow. The MATCH Act threatening $19 billion in chip equipment revenue. The Fed losing its chair during a war. Oil above $100 for the ninth straight week.
If you're a retail investor, you can't control the headlines. But you can control your process. Are you holding things because you understand the cash flows? Or because you're hoping the next headline brings back the crowd?
One company built for the regime where proof matters more than promise:
Company: Automatic Data Processing (SYM: ADP)
The largest U.S. payroll processor — recurring revenue, minimal capex, decades of dividend growth, and a business model that doesn't need AI hype or rate cuts to compound.
ADP is the anti-hype stock. It grows when people get paid. It grows when regulations get complicated. It grows when the economy adds jobs and when employers need help managing complexity. In a regime where the market starts demanding proof — real cash flow, real returns on equity, real balance-sheet strength — ADP is the name that doesn't flinch.
Bottom line: When profit-taking becomes "acceptable" again, the market is quietly changing the rules. Act like it.
Before You Go
Here's the contrarian question for today:
If AI is truly inevitable… why does every earnings season now feel like a negotiation between CEOs and the market over how much spending is "allowed"?
That's not what unstoppable demand looks like. That's what a boom looks like right before discipline returns.
Found this helpful? Share it with others.
Written by Behind the Markets