The largest buyback increase in history is not financial engineering. It is Jensen Huang telling you Q3's $108B guide will be beaten — and that he can fund Rubin, $25B in new debt, and still return $235B by FY28.
The call: BUY NVDA $220–235, target $280–300, invalidation daily close below $195 — 1–3 months, HIGH conviction.
Why now: $150B added to buyback ($235B total through FY28) + Vera Rubin in full production + Q3 guide $108B with zero China — three independent cash-and-demand signals on the same day.
The disagreement: Consensus sees a 5.21% 10Y as a multiple killer; BofA history says multiples hold to ~7%, and NVDA at 28x TTM with 106% revenue growth is the stock that survives 5%+ rates.
The level that changes everything: Weekly close below $195 breaks the post-August earnings structure — otherwise every $205–212 pullback is for sale by the company itself.
NVIDIA authorized an additional $150B under its existing repurchase program, lifting remaining authorization to $235B to be executed through FY28 — the largest authorization increase in history. The stock rose ~2.2% to $230.08 intraday Monday while the S&P fell ~0.76% and the Dow dropped 300+ points on oil and 5.21% 10-year yields. That divergence is the signal.
Three things landed together: the buyback announcement with Huang's "once-in-a-generation platform shift" language; the Open Agent Safety Platform launch with 100+ partners (OpenShell + Sentry, free and open); and Huang on CNBC's Squawk Box saying "we're going through the largest infrastructure build-out in human history" and pledging to return more cash every year. Read the CNBC account of the interview and authorization — the tone is a company guiding up, not defending.
Context matters: this is the third expansion in 13 months — $60B (Aug 2025), $80B (May 2026), now $150B — while H1 FY27 repurchases already ran $39.8B for 203M shares plus $6.29B in dividends, per the 10-Q detail compiled here. A company that just issued $25B in senior notes in June and still accelerates returns is not borrowing to buy back — it is monetizing a backlog.
Q2 FY27 (reported Aug 26, 2026) is the best fundamental quarter any semiconductor company has ever printed — and Q3 guidance already assumes the hardest case. Revenue $96.2B, +18% q/q and +106% y/y; Data Center $89.0B, +117% y/y; GAAP and non-GAAP gross margin both 75.0%; GAAP net income $59.7B; non-GAAP EPS $2.22, +120% y/y. Full primary Q2 release here.
What I weight most from that release:
At $230, NVIDIA screens at ~28x TTM EPS ($7.91) and ~$5.5T market cap — but the run-rate math says ~26x current earnings power with 80%+ growth, which is cheap for a toll asset.
Do the checkable arithmetic with me. Q2 non-GAAP EPS $2.22 annualized = $8.88. At $230 that is 25.9x. Q3 guide $108B at 74% gross with ~$9B opex implies ~$70B operating income — another step up. Even flat Q4 gets FY27 non-GAAP EPS near $9.50–10.00, putting the forward multiple at ~23–24x for a company growing revenue 83% y/y with 75% gross margins and 110% ROE. That is the load-bearing assumption (see appendix): if Rubin gross margin holds ≥73%, the multiple compresses by itself through earnings; if it slips to the mid-60s on Rubin ramp costs, fair value drops ~15%.
The buyback adds a mechanical kicker most models miss. $235B through FY28 is ~4.3% of market cap per year if executed evenly — but NVIDIA repurchased $39.8B in H1 alone. At $220–235, $50–60B/year retires ~1% of shares quarterly, adding ~4–5% annual EPS accretion on top of organic growth. That is why the authorization is price-support, not just signaling: every dip toward $205–212 meets the one buyer with perfect information and no flow constraints.
Against rates: the BofA history — multiples stable at ~16x until the 10Y approaches 7% — matters more for NVDA than for the market. A 28x TTM multiple at 5.21% 10Y looks rich until you realize earnings are doubling; the PEG is ~0.3x on 3-year EPS growth of 204%. The risk is not duration — it is a Rubin delay.
Buy the buyback floor, not the breakout: scale $220–235, target $280–300, kill on a daily close below $195.
| Field | Level |
|---|---|
| Entry zone | $220–235 (scale; add on $205–212 if offered) |
| Target | $280 base / $300 stretch (22–30% upside) |
| Invalidation | Daily close below $195 (breaks Aug-26 earnings gap + 50-day structure) |
| Timeframe | 1–3 months (into Q3 print late November) |
| Conviction | HIGH — earnings acceleration + $235B floor + Rubin ramp = 3 independent signals |
| Audience / sizing | Core large-cap growth; 3–5% position, half now / half on $212–215; hedge with short duration (T-bills), not with a QQQ short |
Why $195: it sits below the $209–212 zone from our prior WATCH-NVDA-1 call (now retired — price blew through it to $230, so that watch resolves as missed entry, thesis confirmed) and below the post-Q2 gap. I never set an invalidation inside my own base case — $195 is a demand-failure print, not noise. Why not chase $236.54 (52-week high, May 14)? Because the buyback gives you the dip bid; there is no edge in paying the high the day the company tells you it will buy every dip for two years.
Base case (55%) is $280: Q3 revenue $110–112B (above $108B guide), Rubin ~$20B+, gross margin ≥74% — stock re-rates toward $280 into the November print.
| Scenario | Weight | Trigger | Price |
|---|---|---|---|
| Bull | 25% | Q3 beat + Q4 guide ≥$120B, Rubin ramp confirmed + China compute relief headlines; hyperscaler capex confirmed >$1.3T | $300–330 |
| Base | 55% | Q3 $108–112B, GM 73.5–74.5%, buyback execution ≥$15B/quarter; 10Y holds 5.0–5.5% | $270–285 |
| Bear | 20% | Rubin supply/CoWoS bottleneck, GM <72%, or sovereign orders pushed; 10Y >6% forces AI-multiple compression | $195–205 |
Second-order effects most coverage misses: (1) the Open Agent Safety Platform (OpenShell isolation + Sentry in-silicon monitoring, backed by SAP/HP/Lenovo/IBM/Cisco statements today) is a CUDA-style moat extension into enterprise governance — it makes Rubin the compliant default for regulated agents; (2) $500B third-party financing platforms move capex risk off hyperscaler balance sheets, extending the buildout even at 5%+ rates; (3) CPU side (Vera) + Groq-3 LPX inference means NVIDIA now monetizes both training and agentic inference — the "compute is revenue" line from the Q2 call is literal.
I am wrong if Rubin economics break or demand concentration snaps — both measurable, neither about narrative.
If none of those fire, hold through the Q3 print. If two fire, exit even above cost — discipline over thesis.
Start half today $228–233, bid the rest $212–215, no chase above $240. Pair with T-bills/short duration (our standing MACRO-DURATION stance), not with hedges that short the AI capex you are long. AMD at $603 (‑4.3% today) and AVGO at $350 are the sympathy reads — but they are sentences, not sections: AMD is the CPU-agentic beta, AVGO the custom-silicon hedge; neither has a $235B floor announced today. Crypto (BTC $83.3k, ETH $2,677, greed 74) can wait a day — this is an equity-cash-flow story.
Prior open call update: WATCH-NVDA-1 ($209–212 watch) is retired — price at $230 confirms the direction but denies the entry. Replaced by EQ-NVDA-2 above. No contradiction with standing EQ-ORCL-1 fade, EQ-MU-1 washout buy, or MACRO-5PCT-1 TLT fade — all share the same regime view: 5%+ 10Y hurts duration, not AI cash flow.
Contents: A. Data snapshot · B. Model & load-bearing assumption · C. Sources
| Item | Level |
|---|---|
| NVDA | $230.08, +2.23% (open $226.51, high $233.21); 52wk $164.27–$236.54; mkt cap ~$5.42T |
| Q2 FY27 | Rev $96.221B (+106% y/y); DC $89.0B (+117%); GM 75.0%; GAAP NI $59.688B; non-GAAP EPS $2.22 |
| Q3 FY27 guide | Rev $108B ±2%; GM 74.0% ±50bp; opex ~$9.2B GAAP / $9.0B non-GAAP; zero China DC compute |
| Buyback | +$150B authorized Sep 28 → $235B remaining through FY28; Q2 returned ~$26B; H1 repurchases $39.8B (203M shares) + $6.29B dividends |
| Balance sheet (Jul 26) | Cash $22.4B + debt securities $34.1B + equity securities $42.8B; assets $320.3B; equity $229.0B; June $25B notes issued |
| Market regime | SPY $765.46 −0.76%; QQQ $735.57 −1.20%; 10Y >5.21% (19y high); SPX ~1.4% below Aug-13 intraday record 7,816.70 |
| Peers / crypto | AMD $603.29 −4.34%; AVGO $350.32 −0.71%; MU $1,045 −3.43%; SNDK $1,696 −4.60%; BTC $83,337 −1.31%; ETH $2,677 −0.45%; Fear & Greed 74 (Greed) |
Q3 $108B × 74% gross = $79.9B gross profit − $9.0B non-GAAP opex = ~$70.9B operating income. Less ~$0.5B net interest drag, tax 17% → ~$58.4B net income ÷ ~24.2B diluted shares ≈ $2.41 quarterly EPS, +8.5% q/q. Annualized run-rate ≈ $9.30–9.60. At $230 = 24x forward. Target $280 = 29–30x forward — justified if growth >50% sustains, in line with 3-yr EPS growth 204% and PEG <0.5x.
Load-bearing assumption: Rubin holds gross margin ≥73%. If Rubin mix + HBM costs drag GM to 68%, Q3 operating income falls ~$6.5B and EPS to ~$2.20, fair value drops to ~$235–245 (no upside). Watch H1 inventory build ($21.4B → $31.6B) and receivables ($38.5B → $63.1B) — if DSOs stretch further in Q3 without revenue conversion, cash quality weakens even if EPS prints.