Nvidia's chip order book doubling from $500B to $1T in four months is not a rumor — it's a structural demand signal that reframes the 35x TTM P/E as a growth discount rather than a premium. Google at 26x TTM P/E with Gemini 3.1 Pro, a $3.48T market cap, and an 18.5x EV/EBITDA is quietly the most undervalued AI platform in the cohort. Both names surged 5%+ on April 1st — institutional rotation is happening in real time.
Let's start with Nvidia, because the order book revision is the single most important data point in Mag 7 coverage this week. Four months ago, Nvidia's combined Blackwell and Rubin chip orders were pegged at $500 billion through end of 2027. That number has now doubled to $1 trillion, implying a booking run rate of approximately $1.3 trillion annually. If you annualize that and apply even conservative margin assumptions, analysts are penciling in $20–25 EPS on near-$1 trillion revenue exiting 2027. At $174.40 — up 5.59% on April 1st — NVDA is trading at a significant discount to that earnings power. The 35.6x TTM P/E looks expensive until you realize the 'T' in TTM reflects a business that was half the size it's about to become.
The Rubin platform catalyst is the next chapter. Production shipments of Vera Rubin chips are expected in H2 2026, and the specs are genuinely disruptive: 10x lower inference token costs versus Blackwell, 4x fewer GPUs required to train mixture-of-experts models. This matters enormously because one of the persistent bear cases on Nvidia has been efficiency gains reducing GPU demand — Rubin flips that narrative. Lower cost per token means more inference workloads become economically viable, which expands the total addressable market rather than cannibalizing it. Sam Altman and Elon Musk publicly expressing enthusiasm for the platform isn't just noise — it signals that the two most aggressive AI compute buyers are already in the Rubin pipeline.
Now pivot to Google, which I think is the most misunderstood setup in the Mag 7 today. At $287.56 — also up 5.14% on April 1st — GOOGL trades at 26.6x TTM P/E and 18.5x EV/EBITDA on $402.8B in revenue. Compare that to Microsoft at 23x P/E on $305.5B revenue, or Apple at 32x on $435.6B. Google is generating more revenue than Microsoft at a cheaper multiple, with an AI product cadence — Gemini 3.1 Pro in February, ongoing Search and Discover core updates — that the market is consistently underpricing. The February Discover update and March 2026 core update are not just SEO events; they represent Google hardening its moat around Search quality at the exact moment AI-native alternatives are trying to disintermediate it.
The valuation case for GOOGL is straightforward: 18.5x EV/EBITDA for the world's dominant search engine, the second-largest cloud platform, a leading AI research organization, and a YouTube advertising business that alone would trade at a premium multiple as a standalone. The market is applying a conglomerate discount and an AI-disruption fear premium simultaneously. I think both are excessive. Gemini 3.1 Pro with 'major upgrades for complex problem-solving' combined with Google Cloud's expanding enterprise AI footprint suggests the company is not ceding ground — it's fighting and investing. The sovereign AI and cloud infrastructure buildout is real, and Google's 20+ years of ML infrastructure gives it a moat that a 26x P/E simply doesn't reflect.
Bringing it together for the portfolio view: both NVDA and GOOGL participated in the April 1st rally with 5%+ moves, suggesting this isn't single-stock idiosyncratic — institutional money is rotating back into AI infrastructure beneficiaries after the Q1 correction. Nvidia remains the highest-conviction growth name in the Mag 7 with the clearest demand visibility ($1T order book, Rubin in H2 2026), and the beta of 2.375 means volatility is the price of admission. Google is the value play hiding in plain sight — lower beta (1.11), cheaper EV/EBITDA than Meta or Apple, and an AI product pipeline that is genuinely competitive. I'm raising overall stance to BULLISH with higher conviction on both names specifically.