INTC +24%. NOW -17%. Same week. The market just changed how it grades AI earnings.

Last Tuesday, two enterprise tech companies reported earnings.

Both beat. Both raised full-year guidance.

INTC closed +23.6%. NOW closed -17.5%.

A 41-point spread. Same earnings week. Same macro backdrop. Not a coincidence.

The market just changed the rules on how it grades AI earnings. Wednesday, four hyperscalers report in a single four-hour window and find out where they stand.

The 3 new rules

IBM and SAP also reported last week. Three of the four beat headline numbers. Only INTC got rewarded.

Why?

Rule 1: Reported numbers, not constant currency.
SAP grew Cloud +27% on a constant-currency basis. Reported revenue missed by $130M. Stock down 4%.

Rule 2: Forward acceleration, not trailing comps.
IBM beat top and bottom and grew its AI book 40%+. CFO on the call: "I don't think we've ever raised guidance in a first quarter." Stock down 8%.

Rule 3: AI revenue must be additive, not substitutive.
NOW's Now Assist ACV grew 50% to $1.5B. The market's read: every dollar of Now Assist is a dollar that would have been spent on legacy NOW anyway. Same wallet. Lower margin. Multiple compression. Stock down 17%.

INTC passed all three. Up 24%.

If you internalize one thing from this newsletter, internalize those three rules. They're the lens for every report between now and August.

Wednesday: $700B in capex hits the tape

Four hyperscalers report calendar Q1 2026 after the close on April 29.

Six hundred billion dollars. Four companies. One calendar year. Roughly Saudi Arabia's annual GDP. Roughly 90% of those four companies' combined operating cash flow.

The market hasn't picked a side on whether the spend pays back. Wednesday is the first quarter where it has to.

Here's how each scores against the 3 rules going in:

GOOGL: cleanest revenue offset to its capex.
Only Mag7 name up YTD (+8%). Cloud +48% last quarter. $240B backlog disclosed at Cloud Next. BMO upgraded April 22 calling it "the best way to own AI." I'll be watching the Cloud growth print: above 40% would confirm the setup, below 35% would break it for me.

AMZN: biggest capex, most visible monetization.
AWS +24% last quarter, fastest in 13 quarters. Bedrock +60% QoQ. Anthropic just committed $100B over 10 years. Capex is the highest at $200B, but the revenue acceleration is the most visible in the bucket. Druckenmiller increased his position 800% last quarter, which tells you how at least one well-known book is reading it.

MSFT: most concentrated counterparty risk.
Down 23% YTD, worst Mag7 stock not named Tesla. Azure decelerating: 40% → 39% → guided 37-38%. 45% of the $625B backlog is OpenAI, who just signed $100B with AWS. That looks like single-counterparty risk to me, in a name the market had been pricing for perfection.

META: highest variance setup.
Cheapest in the bucket at 17x forward. $115-135B capex against zero direct Llama monetization. Druckenmiller sold his entire position in Q4. Ackman bought $1.8B. Both can't be right. I'll be watching for any concrete Llama or Advantage+ ROI number on the call.

How I'm thinking about it

If you own the Mag7 as a bloc (QQQ, MAGS, owning all seven):
The names that look like they pass the filter and the names that look like they fail are inside the same wrapper. The intra-bloc spread (AMZN +8% YTD vs TSLA -19% YTD) is now 27 points wide. The bucket is cancelling itself out. Worth being aware of - what you do about it is a personal call.

If you're in cash or generic large-cap:
GOOGL has the cleanest setup going in. Up YTD, a clean Cloud beat would be the catalyst, downside is bounded by the YTD strength. The asymmetry sits there if you want it.

If you're holding generic SaaS:
Worth running the 3 rules before the print. Three "no" answers and the setup looks asymmetric down. Three "yes" and it looks asymmetric up.

The Tesla canary already broke last Tuesday. Capex guide jumped from $20B to $25B+. Musk explicitly anchored the spend against Amazon ($200B) and Alphabet ($175-185B), saying Tesla was joining the hyperscaler capex bucket. The market replied by treating Tesla like one: stock closed -3.6% on 47% above-average volume.

That's the read-through to Wednesday. Direct.

The bucket framing is doing less work each quarter. The filter is doing more. That's how I'm reading the setup going in.

Nothing in this newsletter is investment advice. It's a journal of how I think about the market. Do your own work.

Charlie