NVDA's $81.6B Quarter Settles the Debate — Data Center Nearly Doubled, Q2 Guides $91B, and Baker's Exit Was Early

The Lead

Issue 07 set a falsifiable test: “if NVDA’s Q1 print shows revenue ≥$50B with Data Center segment ≥40% YoY growth AND guidance maintains $60B+ for Q2, Aschenbrenner’s bet was right.”

On Wednesday 20 May — eight days ahead of our predicted reporting date — NVIDIA delivered Q1 FY27 results. The bars, and the actual results:

Bar (Issue 07)	Actual
Revenue ≥ $50B	$81.6B — cleared by $31.6B
Data Center ≥ 40% YoY growth	+92% YoY — more than double the bar
Q2 guidance ≥ $60B	$91.0B — 50% above the bar

The debate between Aschenbrenner (NEW NVDA $1.57B Q1 2026) and Baker (NVDA cut -77% Q1 2026) is closed on the near-term timing question. Aschenbrenner entered ahead of a generational quarter; Baker exited ahead of a generational quarter. Baker’s structural thesis — that custom silicon will eventually erode NVDA’s margin — may still be correct on a 3-5 year horizon, but the near-term trade was wrong.

One correction before continuing: Issue 07 predicted “NVDA Q1 FY26 earnings on Wednesday 28 May.” Two errors — NVDA calls this quarter Q1 FY27 (their fiscal year ends late January, so FY27 Q1 = Feb-Apr 2026), and the date was May 20, not May 28. The quarter covered was correct; the label and date were wrong. Issue 07’s consensus estimates ($48-52B) also reflected an earlier quarter’s numbers — the street consensus as of May 20 was $78.98B, which NVDA beat to $81.6B.

Next falsifiable form: If NVDA delivers Q2 FY27 revenue ≥$91B (+/-2% guidance range) with Data Center growth ≥20% Q/Q and gross margins ≥74.5%, the thesis arc from Issue 03 (hyperscaler capex $400B+ annualised) is validated for a second consecutive quarter. Earnings expected late August 2026.

Threads

The NVDA verdict: all three falsifiable bars cleared by wide margins

NVIDIA Q1 FY27 (ended April 26, 2026), reported May 20, 2026:

Metric	Q1 FY27	Q4 FY26	Q/Q	Q1 FY26	Y/Y
Revenue	$81.6B	$68.1B	+20%	$44.1B	+85%
Data Center	$75.2B	$62.3B	+21%	$39.1B	+92%
Gross margin (non-GAAP)	75.0%	75.1%	-0.1pts	60.8%	+14.2pts
EPS (non-GAAP)	$1.87	$1.59	+18%	$0.78	+140%
Operating income	$53.5B	$44.5B	+21%	$21.6B	+147%
Net income (GAAP)	$58.3B	—	—	—	+211%

Data Center networking revenue alone was $14.8 billion — a sub-line that didn’t exist as a material number two years ago and is now larger than most semiconductor companies’ total revenue.

The Aschenbrenner position scorecard (Q1 2026 entries, all NEW):

Ticker	Entry basis	Post-print read
NVDA $1.57B	~$130-150 avg Q1 price	Print confirms demand thesis; up ~50-65% from entry
AMD $989M	~$100-120 avg Q1 price	AMD Q1 DC $5.8B (+57%) — share-shift story proceeding
TSM $897M	~$170-185 avg Q1 price	CoWoS allocation 60% to NVDA; TSMC is the picks and shovels
AVGO $1.0B	~$190-220 avg Q1 price	ASIC co-design; verdict on June 3-4 earnings
MU $1.01B	~$100-110 avg Q1 price	HBM supply critical path for NVDA; memory trade reawakening this week
ASML $500M	~$670-730 avg Q1 price	EUV monopoly; all AI chips need ASML equipment
ORCL $1.07B	~$150-170 avg Q1 price	AI cloud and inference infrastructure

Every name in Aschenbrenner’s basket got a directional tailwind from the NVDA print. The $2.05B VanEck ETF position (likely SMH semiconductor ETF) is a basket proxy — the underlying individual names moved.

Baker’s de-risk scorecard:

• NVDA cut -77% (Q1 2026, from ~$950M to ~$218M at Q1 prices) — the quarter he exited delivered +85% YoY revenue
• MSFT exited entirely — not directly implicated in the NVDA print
• QQQ -60% — index hedge that paid off briefly in the tariff selloff but reversed with the earnings-season rally

The single mitigating fact: Baker was not wrong about custom silicon in the abstract. AMD has won 6GW OpenAI and 6GW Meta commitments. Broadcom + Marvell control 95% of ASIC co-design. The custom silicon share-shift is real. He was wrong about timing — NVDA’s volume growth is currently so large that even as AMD takes share, NVDA’s absolute numbers still compound. That may change in FY28-29 as MI400-series and ASIC programs reach scale.

Down 1.5% on a $30B beat — what the stock’s non-reaction means

NVDA’s stock fell approximately 1-1.5% in after-hours trading on May 20 and opened lower on May 21. Multiple analysts described it as a “garden variety beat.” This is worth understanding.

Three explanations, in order of weight:

1. Well-telegraphed by hyperscaler capex. Before NVDA reported, Microsoft, Google, Amazon, and Meta had all disclosed Q1 capex numbers showing massive AI infrastructure spend acceleration. The market had already deduced that NVDA’s revenue would be strong — the print confirmed what the channel had already implied. There was no genuine information surprise.

2. NVDA is 16% of S&P 500 weighting. At $5.4 trillion market cap, NVDA is so large that even a 5-10% move would add or subtract ~$270-540 billion from US equity market cap. The position is already maximum-weight for most funds. There is no incremental buyer who was waiting for a good quarter to finally own NVDA.

3. China exclusion and $50B TAM loss. NVIDIA explicitly stated Q2 guidance assumes no Data Center compute revenue from China. Huang has referenced a $50B China AI chip TAM that is now effectively closed to NVDA by export controls. The $91B guide is a strong number but carries the shadow of a material market NVDA cannot currently access. When/if export controls ease — which Huang suggested is his “sense” over time — that’s the next upside catalyst.

What the non-reaction doesn’t mean: it doesn’t mean the thesis is wrong or that Aschenbrenner was wrong to enter. A stock declining 1.5% on a $30B revenue beat is not “bad news” — it’s the market saying “we already knew this was good; show us what’s next.” The “what’s next” is $91B Q2 guidance and Vera Rubin platform shipments in FY27 H2.

Lab calibration note: the signal here is the 13F positioning (Aschenbrenner NEW $1.57B Q1 2026), which is a strategic_shift × bullish setup. Historical mean alpha at 20d for this setup is +2.84% (n=89). The post-earnings price action (-1.5%) is within noise on that time window; the 20-day alpha from the Q1 13F entry date (before the print) will be measured when we have the price data.

$91B Q2 guidance, no China — the next test and what it implies for the thesis arc

NVIDIA Q2 FY27 guidance (May-July 2026 quarter):

• Revenue: $91.0B ± 2% (range $89.2B–$92.8B)
• Consensus heading into guidance: $86.84B → NVDA’s guidance is ~5% above the street
• No Data Center compute revenue from China assumed
• Gross margins: 74.9%/75.0% GAAP/non-GAAP ± 50bps — stable

What this implies for the Issue 03 thesis (“$400B+ annualised hyperscaler capex”):

NVDA running at $75.2B Q1 Data Center revenue = $300B+ annualised from NVDA’s non-China customers alone. Add AMD’s ~$5.8B Q1 DC ($23B annualised), ASIC programs from AVGO/MRVL ($15-20B annualised combined), and the total AI chip spend comfortably exceeds $350B annually and is still accelerating. The Issue 03 $400B+ annualised thesis was directionally right and may be conservative.

The next three earnings events that test the thesis arc:

Event	Date	What it tests
MRVL Q1 FY27	27 May 2026	Baker’s custom silicon / AI connectivity IC thesis — can MRVL show ASIC revenue inflection?
DELL Q1 FY27	~29 May 2026	AI server revenue — is NVDA chip demand flowing through to system integrators?
AVGO Q2 FY26	3-4 June 2026	Broadcom’s ASIC co-design revenue — the custom silicon share-shift thesis directly tested

MRVL is particularly important for the Baker thesis. Marvell is the second-largest ASIC co-designer (after Broadcom). Baker pivoted hard into Astera Labs (+109%, $369M) and Credo NEW ($102M) — the AI connectivity IC layer below ASIC design. If MRVL’s Q1 FY27 shows ASIC revenue accelerating, it validates that the picks-and-shovels layer (Baker’s new positioning) is capturing value even as NVDA dominates the headline GPU number.

New NVDA reporting structure (effective Q1 FY27):

• Data Center: split into Hyperscale and ACIE (AI Clouds, Industrial, Enterprise)
• Edge Computing: gaming, automotive, professional visualization

This is the right reporting structure for where the business is. The prior Gaming/Data Center split understated how GPU-compute-driven the entire business has become. Watching the Hyperscale sub-line will give more granular read on hyperscaler concentration risk.

Memory trade reawakens, Credo up 8%, NextEra acquires Dominion

Three market signals this week that touch our tracked positions:

The memory trade reawakening. Press this week: SanDisk +9%, Western Digital +5%, Micron +3% in a single session as the memory trade reignited. Context: HBM (High Bandwidth Memory) is a binding constraint on NVDA’s ability to ship Blackwell systems. Micron is NVDA’s primary HBM supplier alongside SK Hynix and Samsung. As NVDA’s $81.6B quarter demonstrates sustained AI chip demand, the downstream HBM demand thesis extends proportionally. Aschenbrenner’s MU $1.01B NEW position — his largest single memory holding — is the direct expression of this thesis. The weekly price action is early validation.

Credo Technology up 8% on Rebellions AI factory partnership. Baker’s Issue 07 NEW $102M Credo position got its first directional print this week: Credo announced an AI factory connectivity partnership with Rebellions, the South Korean AI chip startup. Credo’s high-speed interconnects and PCIe retimers handle the bandwidth between GPUs and the rack-level fabric — exactly the bottleneck Baker identified when he pivoted from NVDA (-77%) toward connectivity ICs. The Rebellions deal is one data point, not a thesis validator, but the direction is right. The structural argument Baker made — that AI connectivity IC demand grows regardless of which GPU vendor wins — is being expressed in live market pricing.

NextEra acquires Dominion Energy for ~$67 billion. This week’s biggest utility transaction: NextEra Energy (NEE) announced an all-stock acquisition of Dominion Energy (D) for approximately $67B, explicitly driven by data center power demand overwhelming Dominion’s Virginia footprint. This is the power infrastructure sub-theme that has been building across our issues — Berkshire’s Delta re-entry (airline fuel efficiency play), Baker’s Vistra NEW ($79M), Tepper’s VST+NRG positions ($304M + $253M = 9.4% of book), and now a $67B M&A confirmation of the theme. The power grid constraint is the real bottleneck in the AI build cycle — not chips, not capital. NEE + D combined would own a dominant share of East Coast grid capacity at the exact moment Virginia data centre park is the world’s densest AI compute cluster.

This is an off-watchlist signal. None of our tracked funds hold NEE or D materially in 13F data. But the M&A premium and the rationale confirm the thesis that Tepper’s utility positioning is the most structurally durable bet in our universe.

Watchlist Updates

Ticker	Direction	Driver	Updated reading
NVDA	bullish (upgraded from conflicted)	Q1 FY27: $81.6B revenue, DC $75.2B (+92%), Q2 $91B guide	Issue 07 debate closed. Aschenbrenner’s thesis confirmed on every metric.
MU	bullish (new signal)	Memory trade reawakening; HBM supply-chain thesis for NVDA demand	Aschenbrenner’s $1.01B MU position getting directional tailwind from NVDA beat
CRDO	bullish (reinforced)	Rebellions AI factory partnership +8%; Baker’s $102M thesis printing	Solo in our tracked universe (Baker only). First directional print is right.
MRVL	watch — verdict 27 May	Q1 FY27 earnings — tests Baker’s custom silicon / connectivity IC thesis	Beat + strong AI revenue guide would partially rehabilitate Baker’s structural thesis even if his NVDA timing was wrong
AVGO	bullish	ASIC co-design leader; Aschenbrenner NEW $1.0B; NVDA print validates AI capex	Earnings June 3-4; Broadcom’s AI ASIC revenue is the cleanest read on custom silicon share shift
AMD	bullish (reinforced)	OpenAI 6GW + Meta 6GW commitments; AMD Q1 DC $5.8B (+57%); Aschenbrenner $989M NEW	Share-shift proceeding; NVDA total is still 13x AMD; volume growth thesis holds
VST / power utilities	bullish (reinforced)	NextEra-Dominion $67B deal validates AI power infrastructure thesis	Tepper VST $304M + Baker VST NEW $79M now have M&A confirmation of power grid thesis

Date Correction (Issue 07)

Issue 07 stated: “NVDA Q1 FY26 earnings on Wednesday 28 May (US after-close).”

Two errors:

Fiscal year label: NVIDIA’s fiscal year ends in late January. The quarter ending April 26, 2026 is their Q1 FY27 — not Q1 FY26. FY26 ended January 2026 and those earnings were reported February 25, 2026.

Reporting date: NVIDIA announced the conference call on April 29, 2026 (before Issue 07 shipped May 17). The correct date was Wednesday May 20. The system’s catalyst database had the date close (May 21 flagged vs actual May 20) but our editorial text had May 28.

The consensus estimates published in Issue 07 ($48-52B) reflected the wrong quarter — the street consensus for Q1 FY27 was approximately $78-79B at time of writing.

The Week Ahead

Date	Event	Watchlist implication
Monday 25 May	US Memorial Day — markets closed	—
Tuesday 27 May	Marvell (MRVL) Q1 FY27 earnings	Baker’s custom silicon thesis tested — ASIC revenue acceleration or flat?
Wednesday 28 May	No major watchlist events	—
Thursday 29 May	CRM, DELL, SNOW, SNPS, ZS earnings	AI SaaS monetisation layer; DELL is an NVDA server proxy
3-4 June	AVGO Q2 FY26 earnings	Broadcom’s AI ASIC revenue — the custom silicon share-shift thesis tested directly

Issue 09 ships Sunday 31 May with: the MRVL verdict on Baker’s connectivity IC thesis, the AI SaaS earnings layer (CRM as Issue 07’s flagged proxy on enterprise AI monetisation), and the full Aschenbrenner-Baker post-mortem now that the primary verdict data is in.

Methodology and Disclosures

Filings Intel Digest is a bona fide financial publication. Nothing here is personalised investment advice. The editor may hold positions in companies discussed; current positions and policy are at /about. All claims are sourced to publicly filed documents or identified press sources. Backtested or historical figures are direction-adjusted and calibrated against the signal lab; calibration is partial — treat any forward statement as a hypothesis, not a forecast.

Frozen as of close-of-trading 21 May 2026 (Thursday). This issue contains no new 13F filings (none filed in the window 14-21 May); signals are drawn from the NVDA 8-K earnings release (filed 20 May 2026) and press sources for Credo, memory trade, and NextEra/Dominion.

Watchlist now tracking 11 funds (9 Tiger-Cub-style growth + Druckenmiller + Berkshire).