Calls › NVDA › Aug 9, 2026

NVDA No clear edge NVIDIA Corporation · Technology

Headline AI-capex growth appears largely priced in, but a capex downturn does not.

Published Aug 9, 2026 · Horizon 3-6 months

Confidence74/100

Entry (close)

$223.96 2026-08-07

Since publish

-2.7% to 2026-10-27

vs SPY

-2.2%

3-month verdict

Closes 2026-11-06

Thesis

Headline AI-capex growth appears largely priced in, but a capex downturn does not. At $223.96, NVDA is 5% below its 52-week high after a 30.3% six-month gain, while FY2027 consensus already requires roughly 82% revenue growth to about $393.9B and FY2028 estimates require another 25%-40%. The stock is not obviously expensive if those estimates hold—about 24.9x FY2027 EPS—but upside over 3-6 months likely requires another material beat-and-raise, strong Blackwell Ultra execution and evidence that gross hyperscaler capex is converting into NVIDIA revenue. The same evidence leaves meaningful downside if guidance merely meets expectations, margins weaken, or custom silicon and infrastructure bottlenecks reduce NVIDIA's capture. That creates event risk rather than a clearly asymmetric setup.

Bull case

Consensus earnings make the valuation less demanding than the $5.42T market capitalization initially suggests. If Q2 results and forward guidance again clear expectations, Blackwell Ultra scales without margin dilution, and Rubin remains on schedule, FY2027-FY2028 estimates can continue rising faster than the multiple compresses. Hyperscaler capex plans still point to growth through 2027, capacity is reportedly absorbed rapidly, and NVIDIA's platform, networking and software position could preserve more share than custom-silicon forecasts assume. In that case, current-year P/E near 25x would not fully price continued upward revisions.

Bear case

The stock already discounts extraordinary execution: approximately 82% FY2027 revenue growth, substantial acceleration to an average $110.6B of quarterly revenue in the second half, mid-70% gross margins and another 25%-40% growth year in FY2028. Gross capex headlines exaggerate NVIDIA's addressable share, while power, interconnection, HBM and packaging constraints can delay GPU deployment. NVIDIA's largest customers are simultaneously developing lower-cost custom inference ASICs, and hyperscaler capex is pressuring free cash flow. At 5% below the high after a 30% six-month rally, an in-line quarter or cautious guide could cause multiple compression even without an earnings-estimate collapse.

What would change our mind

Bullish interpretation is invalidated if Q2 results or the following-quarter guide indicate that FY2027 revenue is no longer tracking toward approximately $393.9B without offsetting margin or EPS improvement.
Bullish interpretation is invalidated if gross margin falls below guidance and management attributes the decline to structural pricing or competitive pressure rather than temporary ramp costs.
Bearish interpretation is invalidated if FY2027 revenue/EPS consensus is raised materially after earnings while the share price remains near current levels, producing clear earnings-led valuation compression.
The demand thesis is invalidated if major hyperscalers cut or defer AI infrastructure budgets, report persistent unused AI capacity, or disclose broad accelerator-order cancellations.
The execution thesis is invalidated if Rubin slips beyond H2 2026 or Blackwell Ultra supply and rack-level integration issues prevent the required second-half revenue ramp.
The share-retention thesis is invalidated if disclosed customer deployments show custom ASICs or AMD replacing, rather than supplementing, NVIDIA at scale in core training or inference workloads.

Since publication vs SPY · closed horizons are frozen

1 weekfinal

+0.1%

+0.5% raw

1 monthfinal

+3.2%

+2.9% raw

3 monthsrunning

-2.2%

-2.7% raw

6 monthsrunning

-2.2%

-2.7% raw

The full research everything the view rests on

Where the edge would come from

What the price assumes

The price appears to assume continued rapid AI infrastructure deployment, sustained NVIDIA dominance, roughly $393.9B of FY2027 revenue and $9.00 of EPS, gross margins around the mid-70% area, and another large growth year in FY2028. It also implicitly assumes that hyperscaler capex growth through 2027 converts into accelerator and networking purchases without a major product delay, power-driven deployment pause, or material loss to custom ASICs. Announced capex itself is therefore not differentiated information; NVIDIA's capture rate, quarterly conversion and forward guidance are what remain under debate.

Why the market may be wrong

The bearish potential error is treating total capex as NVIDIA-addressable demand: budgets also fund buildings, power, cooling, networking, CPUs, storage and internal accelerators, and one outside estimate places GPU/CPU spend near 20% of AI-infrastructure capex. The bullish potential error is anchoring on deceleration before it appears in orders: Q1 FY2027 grew 85.2%, Q2 guidance exceeded the then-current Street estimate, hyperscaler budgets have continued to rise, and deployed AI capacity reportedly remains absorbed quickly. The decisive question is not whether capex grows, but whether NVIDIA can sustain an unexpectedly high share of that growth while preserving roughly 75% gross margins.

Key evidence (10)

NVDA closed at $223.96 on August 7, 2026, only 5% below its 52-week high; it had risen 30.3% over six months and 10.4% over one month, with approximately 40%-43% annualized 20-day volatility (computed market data, August 7-9, 2026).
NVIDIA reported FY2026 revenue of $215.94B, Data Center revenue of $167.45B, diluted EPS of $4.90 and gross margin of 71.1% (NVIDIA FY2026 Form 10-K, filed February 25, 2026).
Q1 FY2027 revenue was $81.62B, up 85.2% year over year; EPS was $1.87 and gross margin was 75.0%. Management guided Q2 revenue to $91B plus or minus 2%, above the then-Street estimate of approximately $86.84B (NVIDIA Q1 FY2027 release, May 20, 2026).
Public FY2027 consensus was approximately $393.85B of revenue, $9.00 of EPS and $212.9B of FCF as of August 9, 2026. Relative to FY2026, that implies about 82% revenue growth and values the stock at roughly 24.9x EPS and a 3.9% prospective FCF yield.
Q1 actual revenue plus the Q2 guidance midpoint equals $172.6B. Reaching FY2027 consensus requires approximately $221.2B in the second half, or $110.6B per quarter on average, making continued sequential expansion a factual requirement rather than optional upside.
Public FY2028 estimates vary materially but cluster around $492B-$552B of revenue and $10.74-$12.89 of EPS, implying another approximately 25%-40% revenue growth year and a current valuation around 17.4x-20.9x FY2028 EPS.
Reuters reported UBS estimates that hyperscaler capex would grow 76% in 2026 to $673B, 25% in 2027 and 6% in 2028 (Reuters, July 17, 2026). Other methodologies place 2026 spending across Microsoft, Amazon, Alphabet, Meta and Oracle around $700B-$800B.
Blackwell Ultra was reportedly shipping in volume by mid-2026, while NVIDIA's roadmap targets Rubin for H2 2026 and Rubin Ultra for H2 2027. CoWoS packaging and HBM availability remain identified supply constraints; Rubin timing is still roadmap evidence rather than proven shipment evidence.
Industry estimates place NVIDIA at roughly 75%-80% or more of accelerator revenue share, while hyperscaler custom ASICs and AMD are expected to take some share. The exact future share loss is an assumption, not an established fact.
Evidence on capex sustainability is mixed: cloud AI demand and backlogs remain strong, but capex is pressuring hyperscaler free cash flow and depreciation will rise. The evidence supports deceleration after 2027 more clearly than an imminent cancellation cycle.

Catalysts and risks

Catalysts

Q2 FY2027 earnings and guidance scheduled for August 26, 2026 after market close; the critical test is whether revenue, gross margin and the next-quarter guide exceed already elevated expectations.
Evidence that second-half quarterly revenue is tracking toward or above the approximately $110.6B average required by FY2027 consensus.
Blackwell Ultra volume-ramp updates, including rack deployments, supply availability and gross-margin effects.
Rubin launch and shipment evidence in H2 2026 versus the current roadmap timetable.
Further hyperscaler capex revisions and commentary on AI-capacity utilization, backlog and returns.
Changes to U.S. export controls or China shipment approvals that expand or reduce NVIDIA's serviceable market.

Risks

A revenue beat accompanied by guidance that merely matches consensus could disappoint because the current price already embeds rapid sequential growth.
Gross margin falls materially below the approximately 75% level due to system complexity, HBM costs, packaging yields or product-transition expenses.
Hyperscaler custom ASICs capture more inference workloads than expected, reducing NVIDIA's revenue share or pricing power.
AMD gains material anchor-customer volume and forces lower accelerator pricing.
Power and grid-interconnection delays prevent announced data-center budgets from becoming near-term GPU deployments.
Hyperscaler free-cash-flow pressure or weak AI monetization causes 2027 capex plans to be deferred.
China and broader export controls eliminate addressable demand or create inventory and product-design costs.
High short-term volatility—approximately 40%-43% annualized—makes the 3-6 month outcome heavily dependent on earnings-event execution.

What to watch

Q2 revenue versus the $91B guidance midpoint and the magnitude of the next-quarter revenue guide.
Whether full-year consensus still requires an implausibly steep fourth-quarter ramp after the next report.
Data Center compute versus networking growth and any disclosure on Blackwell Ultra mix.
Gross margin trajectory relative to approximately 75%, including product-transition and supply-chain costs.
FY2027 and FY2028 revenue/EPS revisions immediately after earnings, separated from any change in the valuation multiple.
Hyperscaler capex revisions, but especially the portion directed to accelerators rather than facilities, power and proprietary silicon.
Comments on CoWoS, HBM, rack deployment timing and power availability.
Concrete Rubin shipment milestones rather than roadmap reiteration.
Custom-ASIC deployment volumes at Google, Amazon, Microsoft and Meta, plus AMD customer wins.
China shipment assumptions included in guidance.

Market snapshot when the call was made 2026-08-07

Price

Return 1w	+11.6%
Return 1m	+10.4%
Return 3m	+5.9%
Return 6m	+30.3%
Return 1y	+24.8%
From 52-week high	-5.0%
Realized vol, 20d	42.5%
SPY 3m / 1y	+5.7% / +22.2%
QQQ 3m / 1y	+4.2% / +28.1%

Technicals

NVDA technicals as of 2026-08-07: last close $223.96; 20d SMA 206.82 (neutral); 50d SMA 206.07; 12d EMA 210.21; 26d EMA 207.15; RSI 14 65.6 (balanced); 20d vol 39.8%; volume x0.8 vs 20d avg

Next earnings: 2026-11-17 AMC

Sources (76 · T1 = primary filings)

Earlier and later calls on NVDA

Aug 9, 2026AI-capex growth is substantially, but not completely, priced in.No clear edge

Want a fresh look at NVDA, or another name?

Queue a new deep-research run. Published to the feed when done.

Automated research opinion, published as-is on Aug 9, 2026. Not personalised advice. Figures use end-of-day closes.