Amazon's $3 Trillion Moment: AWS Acceleration Real, But Don't Chase

Amazon's stock exploded 15.3% after Q2 earnings, then hit an all-time high on Monday August 3rd and pushed market cap past $3 trillion. Citigroup raised its target to $350. Roth MKM to $325. Wall Street's consensus is 39 buys, 1 hold, 0 sells.

The question isn't whether Amazon delivered. The question is whether the market is celebrating the right thing — and whether the $220 billion capital expenditure plan hiding behind those headlines justifies the enthusiasm or undermines it.

I've been looking at Amazon's numbers since the report landed on July 30th, and the verdict is more nuanced than the post-earnings euphoria suggests. The AWS story is real. The capex burn is real. The one-time accounting gain that made net income look spectacular is also real — and it matters for anyone trying to value this stock on forward earnings.

The Earnings: Strip Away the Accounting Event

Amazon's Q2 net income was $62.6 billion — a 245% year-over-year jump that made the $5.75 reported EPS look like an out-of-body experience for a company that reported $1.68 per share a year earlier. But $53.4 billion of that net income was a non-operating mark-to-market gain on Amazon's Anthropic stake, reflecting higher valuations in recent financing rounds. It's paper appreciation, not recurring cash flow. It's not even recurring revenue.

Strip it out, along with a smaller $600 million energy-derivative gain, and the underlying, operations-driven net income comes to roughly $20.8 billion. That's a more modest 14-20% increase over the prior year. Still solid. Not the generational profit explosion the headline number implies.

The adjusted EPS that actually reflects the operating engine — $1.97 — beat the $1.82 estimate, but that beat is thin. What's more important is what drove it: operating income grew 43% to $27.5 billion, with AWS contributing $16.6 billion of that. AWS accounts for 21% of Amazon's revenue but 60.5% of its operating income. That operating leverage is the real story.

AWS Is Actually Accelerating

AWS revenue hit $42.2 billion, up 37% year-over-year, beating the consensus estimate of roughly 31%. That's the fastest AWS growth in 18 quarters — the unit hasn't grown this fast since 2021. Operating margin expanded to 39.4%, up from 35.1% a year earlier. At a $169 billion annualized revenue run rate, AWS is approaching the scale that turns every basis point of margin expansion into billions of dollars of incremental profit.

What's fueling this? Both the AI business and the chips business (Trainium and Graviton custom silicon) have individually surpassed $25 billion annual revenue run rates, growing at triple-digit percentages. AI workloads aren't replacing traditional cloud consumption — they're driving demand for storage, vector databases, and traditional compute alongside inference. And the $496 billion AWS backlog, contracted work not yet online, is growing at triple-digit rates with demand already visible into 2028.

This isn't a narrative. It's a backlog.

The Valuation Question

Now for the uncomfortable part. Amazon's forward P/E is 42.0x. That's not cheap. The trailing P/E of 21.9x looks benign, but it's distorted by that one-time Anthropic gain. On a forward basis, the market is pricing in sustained AWS margin expansion, advertising growth (up 26% to $19.8 billion in Q2), and continued retail efficiency gains.

The PEG ratio of 0.25 is the number that keeps contrarian investors awake at night. At that level — forward P/E divided by revenue growth rate — Amazon looks dirt cheap relative to its growth trajectory. Revenue grew 20% in Q2, with AWS growing 37%, advertising 26%, and North America retail 16%. The company's revenue growth rate dwarfs the S&P 500's, yet the forward multiple hasn't reached absurd territory because the earnings denominator is also expanding fast.

Compare that to Microsoft at 27.8x trailing P/E with Azure growing 43%, and Alphabet at 17.8x with Google Cloud surging 82%. Amazon's valuation sits between them, which makes sense: its growth is solid but not Google's explosive small-base numbers, and its valuation hasn't stretched to Microsoft's premium.

The valuation isn't the problem here. The execution risk is.

The $220 Billion Capex Problem — And Why It Might Be the Answer

Here's the number that should give cautious investors pause. Amazon's trailing twelve-month free cash flow swung to a $7.6 billion outflow, down from an $18.2 billion inflow a year earlier. TTM operating cash flow grew 33% to $161.4 billion — impressive — but capital expenditures ballooned to $173 billion over the trailing twelve months.

On the earnings call, CEO Andy Jassy raised 2026 full-year capex guidance to $220 billion, up from the $200 billion he projected in February. The increase isn't from building more data centers — it's from higher memory costs, specifically high-bandwidth memory (HBM) for AI accelerators. Supply volatility in memory chips is squeezing margins at the infrastructure layer.

That's a lot of cash going out the door. Total debt has risen to $544 billion, though net debt is manageable at roughly $6 billion thanks to $78 billion in cash and equivalents. The debt-to-equity ratio of 23% isn't alarming. But $220 billion in annual capex is an enormous commitment, and history is littered with technology companies that front-loaded infrastructure spending and never saw proportional returns.

The reason Amazon's bet looks different: Jassy was explicit on the call that even at $220 billion, "we will still not have enough capacity to meet all the demand we have in 2026." He said capacity constraints will persist through 2027, with striking demand already present for 2028. The $496 billion backlog isn't a forecast — it's contracted, multi-year commitments from enterprises and AI labs including Anthropic (which signed a ten-year, $100 billion Trainium commitment in April 2026) and OpenAI.

Jassy also broke down the ROI math: data centers take roughly two years to generate revenue but remain productive for over 30 years. Servers and networking equipment break even in under three years and generate substantial free cash flow before replacement. By his calculation, AWS could "very possibly become a trillion-dollar annual revenue business."

That last claim is aggressive. But the demand visibility behind it — the backlog, the multi-year commitments, the sustained pricing power in a capacity-constrained market — makes it defensible. The capex isn't wasted spend. It's capacity construction in an environment where supply is the bottleneck, not demand.

The risk is that demand softens, memory costs keep rising, or competitive pressure from Microsoft and Alphabet erodes AWS's margin advantage. But right now, the evidence points the other way: all three hyperscalers are accelerating cloud growth simultaneously, suggesting the AI infrastructure cycle is broad-based rather than zero-sum.

Price Action: Momentum, Not Exhaustion

After the 15.3% post-earnings surge, the stock hit $287.20 on August 3rd — its 52-week high and the first time market cap exceeded $3 trillion. The stock is trading at $275, roughly 11% above its 50-day moving average of $247.52 and 16% above its 200-day average of $236.54. The RSI sits at 64.5 — elevated but not overbought. MACD is positive at 7.6, confirming upward momentum.

The stock is up 38% over the past 120 days and 19% year-to-date. This isn't a battered growth stock being mispriced by market panic. This is a stock that just confirmed its thesis with a blowout quarter and is being bid higher as a result.

The price action doesn't signal a bear trap or selling exhaustion. It signals conviction. And while the short-term extension means the risk/reward on a fresh entry isn't as attractive asit would be on a pullback, the trend is your friend when the fundamentals are this strong.

The Action

Here's where I stand.

The market isn't misreading Amazon's fundamentals in the wrong direction — it's responding correctly to genuine AWS acceleration. But there's still a gap between what the stock reflects and what the backlog implies. The $496 billion in contracted demand extending into 2028 is a visibility moat that few growth stocks can match. The 37% AWS growth rate is real, not narrative-driven. The capex spending is aggressive but justified by that same backlog.

I'm reiterating a Buy on Amazon, but with a practical caveat: don't chase.

The stock has already moved 15% off the earnings report and hit all-time highs. The forward P/E of 42x, while defensible given the growth trajectory, is not dirt cheap. The better risk/reward for new positions is likely on a pullback toward the 50-day moving average in the $245-$250 range, or if broader market volatility creates a dip below $260.

For existing holders, the setup supports doubling down on weakness rather than trimming. The AWS acceleration, the backlog, and the margin expansion trajectory all point to continued earnings growth that should compress that forward P/E over time, even if the stock price stays flat.

The Anthropic gain was a windfall, not a pattern. But it signals something important: Amazon's AI investments are creating real value, whether through ownership stakes or through AWS consumption from the companies it backs.

I would reassess if AWS growth decelerates below 25%, if the capex plan continues to expand beyond $220 billion without corresponding backlog growth, or if the stock breaks below the 200-day moving average on sustained volume — that would suggest the momentum thesis is cracking.

Until then, the numbers argue for patience, not panic. The market is celebrating a genuine inflection in AWS. The question is whether you're comfortable with the entry price.