The Lyceum: Follow the Money — Apr 28, 2026

Three things are true at once this week, and they're more interesting in combination than apart. Big Tech is about to defend roughly $645 billion in 2026 AI capital spending in front of an audience that has finally stopped clapping on faith. Insiders at quieter, less glamorous names are buying with their own cash while CEOs at the obvious winners are distributing in size. And Thursday lands a Q1 GDP print, the ECB decision, and Apple earnings on the same calendar square. Match the signals to the moment: this is an accountability week, not a victory lap.

The most important question in markets right now isn't about rates or oil. It's whether the largest corporate capital spending cycle in history is actually working.

Microsoft, Alphabet, Meta, and Amazon all report Wednesday after the close, with Apple following Thursday. Per Bridgewater's estimate cited in Invezz, the four hyperscalers have committed roughly $645 billion to 2026 infrastructure — up about 56% from 2025's $410 billion. The numbers underneath have been constructive: of the 138 S&P 500 members who had reported through April 24, earnings were tracking +14.5% year-over-year on +9.7% higher revenues, according to Zacks. The five Magnificent Seven names reporting this week have rallied more than 10% this month; the Nasdaq is up 15% in April.

What changes if Big Tech delivers: the AI-capex beneficiaries — power producers, custom silicon, data center REITs, copper — get a second leg up as the spending narrative converts into earned revenue. What failure looks like: even one hyperscaler trimming 2026 capex guidance. The signal to watch isn't headline EPS. It's Azure growth, Meta's ad margins, and the Google Cloud backlog (last reported at $243 billion). Those three numbers tell you whether $645 billion is compounding or just accumulating.

Microsoft is on track to spend close to $146 billion on AI and cloud infrastructure in fiscal 2026, against a stated AI revenue target of $25 billion — roughly 17 cents of revenue per dollar of spend. That ratio is the entire conversation.

Wedbush calls fiscal 2026 an inflection point as more CIOs accelerate AI deployment, but the technicals tell you the market hasn't pre-celebrated: Microsoft closed Thursday at $415.75, well below its 52-week high of $555.45, per Investing.com. The single most scrutinized line will be Copilot adoption inside enterprise Office 365.

If Copilot is sticking, the spending math holds and the picks-and-shovels trade extends. If adoption is tepid, the questions get uncomfortable fast — not just for Microsoft, but for every hyperscaler that has used Microsoft's Copilot trajectory as cover for its own capex. Watch commercial bookings and the deferred-revenue line; that's where adoption shows up before management has to characterize it.

Most coverage Wednesday will fixate on Alphabet's search revenue. The more interesting number is buried in the cloud segment.

Google Cloud grew 47.8% in Q4 2025, putting the segment at a roughly $70 billion annual run rate, with a backlog that hit $243 billion. Per Zacks, that Q4 growth was an acceleration from Q3's +34%, and management has guided continued momentum into Q2. Wedbush thinks the Street is still too conservative, citing Gemini 3, custom TPU chips, and the cloud business specifically.

The bull case: cloud growth keeps accelerating, the backlog converts predictably, and Alphabet becomes the cleanest AI infrastructure compounder in mega-cap. The bear case: 47 of 55 analysts already rate it Strong Buy with a $378 consensus target, per HeyGoTrade — meaning the bar for a positive surprise is genuinely high. The signal isn't whether cloud beats; it's whether deceleration appears anywhere in the backlog conversion.

No major S&P 500 company enters earnings cleaner than Meta. Across 42 analysts: zero Sell ratings. Wall Street expects Q1 2026 revenue around $55.5 billion and EPS of $6.65, with an average target of $855.60, per Invezz.

Then there's the other half of the story. Meta has laid off 8,000 employees while maintaining 2026 capex guidance of $115–$135 billion. Microsoft is offering buyouts. Per Yahoo Finance, this earnings cycle will give investors their first real read on whether companies are reconsidering AI spending, and Wedbush expects Meta to demonstrate the cleanest example of AI translating into revenue.

If Meta delivers: the thesis that AI replaces the labor it eliminates gets its first quantitative endorsement, and the headcount-down/capex-up trade becomes a template. If margins disappoint: the layoffs read as defense, not redeployment, and every other hyperscaler's capex commitment gets a second look. The number to track is operating margin in the Family of Apps segment — that's where labor savings and AI monetization meet.

A quiet Friday development may matter more than any earnings line this week. Reports in Yahoo Finance said the Justice Department dropped its criminal investigation into Jerome Powell over Federal Reserve building renovation cost overruns. Per Yahoo Finance, the development could improve Kevin Warsh's political prospects for leading the Fed. LPL Financial chief economist Jeffrey Roach noted the episode "underscores broader concerns about the politicization of the central bank."

What changes: a cleaner Warsh transition path, against a backdrop of headline inflation rising on energy, core PCE still above target, and Q4 2025 GDP at 0.5%, per Kraken's economic brief. What to watch: BlackRock has stayed underweight long-dated Treasuries, citing what it calls a "diversification mirage" as term premium rises. A hawkish Warsh would validate that posture. The signal is the long end of the curve — if 10-year yields drift higher even as growth data softens, the bond market is telling you Warsh is being priced in already.

Between January 22 and April 22, 2026, Palantir co-founder Alexander Karp acquired roughly 1.47 million shares — the largest insider buy in any U.S. growth stock over that window, per HeyGoTrade. The purchases came alongside heavy selling from other Palantir insiders, which is what makes Karp's accumulation legible as conviction rather than program.

The SEC's transaction code on Form 4 matters here. Code P is an open-market purchase — fresh cash, market price, no pre-arrangement. Code A is grants. Code M is option exercises. 10b5-1 plans are scheduled. Insiders sell for many reasons; they generally buy for one. Karp's filings are code P, sustained over four months.

What this is: a co-founder making a sustained personal bet on the AI adoption ramp while the rest of his C-suite cashes chips. What to watch: a technical breakout above recent resistance would complete the convergence and likely trigger institutional momentum buying. Until then, this is the loudest single-name Delta signal on the board.

Small in dollars, loud in signal. On April 23, Travel + Leisure Co. CFO Erik Hoag bought 1,000 shares at $65.67 — about $65,670 — opening a position from zero, per Meyka's reading of the Form 4. Not adding to an existing stake. Starting one.

A CFO has the cleanest forward view of bookings, cancellations, and liquidity at a vacation-ownership business. Travel + Leisure is also a direct consumer-confidence proxy and a name that benefits meaningfully if the Strait of Hormuz reopens and international travel demand normalizes. The transaction code is P — open market, personal capital.

What to watch: a second insider buy at TNL within two weeks would convert this from a single-data-point signal to a clustered one. Right now it's one filing, but it rhymes with the broader pattern this week — selective insider conviction in quieter, beaten-down names while leadership at hot tech distributes in size.

On April 24, eight insiders at Eaton Vance Senior Floating-Rate Trust (ticker: EFOR) bought in the open market simultaneously, including CEO Theodore Hanson at roughly $1.0 million, per Form4Alpha's aggregation of SEC filings. Synchronized open-market buying at a floating-rate credit vehicle is the kind of cluster that usually precedes a story the rest of the market hasn't simplified yet.

On April 14, Meta committed to deploying 1 gigawatt of custom MTIA chips built with Broadcom on a 2-nanometer process, scaling to multiple gigawatts by 2027, per HeyGoTrade. If MTIA performs at scale, every hyperscaler's GPU dependency math gets revisited — and Meta will have set the template.

Per SetYourStop, institutions bought $3.06 million in call premium and collected $2.59 million selling puts — a structure that profits only if the stock holds or rises. One data point; together with other flows, it can be a pattern worth tracking.

A co-founder buys 1.47 million shares of his own company while his colleagues head for the exits, a CFO opens a vacation-timeshare position from literally zero on the same week eight credit-fund insiders pile in together, and somewhere in Washington a Justice Department inquiry into the Fed chair was reported closed so a transition path looks cleaner. Big Tech will spend Wednesday explaining $645 billion to a room that has finally stopped applauding — and the most interesting money this week wasn't anywhere near the microphone.

Stay skeptical.

Forward this to the friend who keeps asking what the smart money is actually doing.