Amazon Proved AI Capex Pays. Meta Is Still Asking for Patience.
Two hyperscaler prints, 24 hours apart, drew opposite verdicts from the market. The difference isn't about who spends more. It's about who can point to the revenue.
Two of the biggest AI spenders in the world reported earnings 24 hours apart this week. Meta went first and got a 7% selloff. Amazon went second and surged 14% in after-hours trading. The difference is not about the size of the capex checks. Both are enormous. The difference is that Amazon can point to the revenue and Meta cannot, not yet.
Q2'26 AMZN FCF not yet available; META Q2'26 = $784M per transcript.
Meta's free cash flow collapsed from $13.2 billion in Q1 2026 to $784 million in Q2 — on $60.8 billion in revenue. That is not a typo. The company is burning nearly every dollar of operating profit on AI infrastructure. Amazon's FCF is also deeply negative, but the operating income line tells a different story: $27.5 billion in Q2, up 43% year over year, with AWS margins expanding to 39.3%.
The AWS Acceleration Nobody Saw Coming
Amazon's Q2 was the best AWS print in 18 quarters. Revenue hit $42.2 billion, up 37% year over year and accelerating for the fifth straight quarter. The quarter-over-quarter revenue add of $4.6 billion was 80% bigger than the previous record.
CEO Andy Jassy said AI margins are tracking "a little bit ahead" of where core cloud margins were at the same stage, directly refuting the thesis that AI workloads are structurally lower margin. The demand signals are hard to argue with:
- AWS backlog hit $496 billion, growing triple digits year over year
- Both Anthropic and OpenAI are now locked into multi-year, multi-gigawatt Trainium commitments
- AI revenue run rate crossed $25 billion, growing triple digits
- The custom chips business (Trainium and Graviton) is also at a $25 billion run rate
- Capex raised to roughly $220 billion, and Jassy says the company still cannot meet all demand in 2026
Meta's Capex Is Eating the Business
Meta's Q2 was a mirror image. Revenue of $60.8 billion beat consensus and grew 28% year over year. The core ads business is a machine: 82% gross margins, Advantage Plus at a $75 billion annual run rate, and AI-driven ad ranking improvements driving measurable conversion lifts. But GAAP operating income fell 8% year over year to $18.8 billion, dragged down by $3.6 billion in one-time charges. EPS of $6.18 missed the $7.17 consensus by nearly a dollar.
Meta now spends 51 cents of every revenue dollar on infrastructure. Amazon spends 22 cents. Both are in a buildout, but Amazon's capex is chasing demand that already exists. Meta's capex is creating demand that doesn't exist yet.
CEO Mark Zuckerberg laid out four AI monetization lanes: core ads, personal agents, enterprise API, and compute rental. But only core ads is generating material revenue today. Personal agents are "shipping soon." The enterprise API is "encouraged by initial results." Compute rental is "getting a lot of offers at a significant premium." These are all reasonable paths, but they are paths, not proof.
The Strategic Fork: Platform vs. Vertical
The strategic contrast is even sharper than the financial one. Jassy argued that Amazon does not need its own frontier model to win: "There is not going to be one model to rule the world." Bedrock hosts every model. Trainium hosts every lab. Amazon collects rent on the entire AI economy regardless of which model is best.
Zuckerberg is betting the opposite: that Meta must be on the frontier, and that the company's 3.6 billion users give it a distribution advantage no other lab can match. Meta's April pivot from open-source Llama to proprietary Muse Spark means it is now competing directly with Anthropic and OpenAI for the same enterprise API dollar — except those two are paying Amazon for compute while Meta builds its own.
The valuation gap reflects the proof gap. Amazon trades at 24 times forward earnings with a PEG of 1.2. Meta trades at 14.7 times forward with a PEG of 0.85. Consensus price targets imply 38% upside for Amazon and 53% for Meta. The Street is telling you the same thing the earnings calls told you: Amazon's AI revenue is visible today, and Meta's is not. Whether that discount is an opportunity or a trap depends on whether Meta's four monetization lanes start producing revenue before the capex bill comes due.
The next checkpoint is Meta Connect on September 23, where the company is expected to detail the Muse Spark API, personal agent launch, and enterprise go-to-market. For Amazon, the story is simpler: watch the AWS backlog conversion rate and whether the acceleration has further to run.
Sources: Amazon Q2 2026 earnings call transcript (July 30, 2026). Meta Q2 2026 earnings call transcript (July 29, 2026). Analyst consensus data from Yahoo Finance via Bargo. Price data from Alpaca SIP via Bargo. SEC filings via EDGAR.
More research at bargo.ai/research.