Phase Space Research

Advanced Micro Devices

AMD · Investment summary · as of 12 August 2026

Priced for an operating path the record does not support

Portfolio decision
No position
Price · 12 August 2026
$474.32
12-month target
$376.00 -21%
Expected return
-10.3%
Next decision point
26 December 2026Thesis condition tested

Business type: Cyclical · cyclical or commodity-sensitive

The operating path required by today's price is not achievable on the evidence.

Investment view

At $474.32, AMD requires a 71% five-year revenue growth rate to justify its enterprise value — more than the business has ever demonstrated, at 24%.

The conditions that would settle the disagreement are dated to December 2026.

The value rests on an exit multiple of 9.8x, a terminal operating margin of 22% and a 11.8% cost of capital. Move any one of them materially and the conclusion moves with it, which is why the required-versus-demonstrated test above carries more weight here than the point value.

The strongest argument against this view: The MI-series accelerator franchise fails to establish a durable merchant share position and the loss is PERMANENT, not cyclical.

Underwriting bridge

QuestionEvidence-based conclusion
What drives the business?State B cyclical/capex-driven; cycle position dated per the State B requirement: all-time-high TTM revenue ($41,305m) and Data Center mix (58%), TTM GAAP operating margin 15.71% in the top decile of AMD's own 19-year filed history and rising, unconditional purchase commitments at an all-time high ($30.3bn)
What do we forecast?Revenue growth of 24% demonstrated; a terminal operating margin of 22%; an exit multiple of 9.8x.
What does Street forecast?Not determined — no consensus figures are on file
Where do we differ?On revenue growth, the difference between what the price requires and what the business has demonstrated is -47.1 percentage points.
What is it worth?Twelve-month target $376.00, -21% from the struck price. Scenario-weighted expected return net of costs -10.3%.
Why now?The first dated test of the thesis falls on 26 December 2026.

What must go right

  1. By 26 December 2026Data Center segment revenue at or above $7,000m in a single quarterWhere it stands: Not determined — no current reading of this metric is on file
  2. By 26 December 2026Remaining performance obligations still at or above $2,000mWhere it stands: Not determined — no current reading of this metric is on file
  3. By any two consecutive fiscal quarters through FY2027Data Center segment operating margin at or above 20.0% in two consecutive quartersWhere it stands: Not determined — no current reading of this metric is on file

Catalysts and falsifiers

Date or windowEventThesis confirmed ifThesis weakened or refuted if
26 December 2026Data Center segment revenue below $7,000m in a single quarterData Center segment revenue at or above $7,000m in a single quarterData Center segment revenue below $7,000m in a single quarter
any two consecutive fiscal quarters through FY2027Data Center segment operating margin below 20.0% in two consecutive…Data Center segment operating margin at or above 20.0% in two consecutive quartersData Center segment operating margin below 20.0% in two consecutive quarters
any two consecutive fiscal quarters through FY2027GAAP gross margin below 50.0% in two consecutive quartersGAAP gross margin at or above 50.0% in two consecutive quartersGAAP gross margin below 50.0% in two consecutive quarters
26 December 2026Remaining performance obligations still below $2,000mRemaining performance obligations still at or above $2,000mRemaining performance obligations still below $2,000m

Risk and sell discipline

Impairment case

The MI-series accelerator franchise fails to establish a durable merchant share position and the loss is PERMANENT, not cyclical. NAMED MECHANISM (primary) - share loss to NVIDIA and to customer custom silicon: OpenAI ('up to 6 GW'), Meta ('up to 6 GW') and now Anthropic ('up to 2 GW', announced with Q2 FY2026 results) are warrant-linked deployment frameworks, not purchase commitments - RPO is $222m (0.54% of TTM revenue, down from $264m in the quarter the partner list grew), and the 10-K cancellation language stands. All three counterparties (or their partners) are simultaneously developing custom accelerators. If the next NVIDIA generation or customer ASICs take the sockets, AMD does not lose a contract; it loses a design-in it never contractually had.

Estimated probability 25%, against the 13% level at which the position would be resized. It sits above that level, so this case could not be carried at full size.

Fundamental invalidation

Falsifiable and fundamental — not one of them is a price condition.

Price-based risk trigger

A daily close below $362.59 triggers an immediate review of the thesis and pauses additional buying. It is not an automatic sell unless a separate fundamental invalidation condition has been breached.

Upside sell discipline

On approach to the $376.00 target the case is reviewed rather than added to; a target reached is a reason to re-examine the position, not to hold it by default. The position is trimmed once forward expected return falls below 0% net of costs, because the capital has a better use elsewhere in the book.

Investment criteria

CriteriaStatusInvestment meaning
QualityMetCYCLICAL test: through-cycle gross margin > 20%.
ValuationNot metIs the operating path required by today's price achievable?
LiquidityMet20-day ADV ~$15.5bn; size is not a constraint at any book scale. Equity is the default rung and no argument for a step above it is offered because no position is taken.
DownsideMetThe MI-series accelerator franchise fails to establish a durable merchant share position and the loss is PERMANENT, not cyclical.
MomentumNot determined12-1 momentum (t-252 to t-21) = +209.3% vs SPY +17.6% over the same window; 12-month total +174.6%; 52-week range $151.14-$580.91, spot at 81.7% of the high.
CatalystMetWindow from AMD's own filed cadence: Q3 10-Q filing dates 2017-2025 run 1 November 2017 to 5 November 2025; the Q3 FY2026 report (period ending 26 September 2026) is expected late October / early November 2026.
ConsensusNot determinedFY2026 revenue $49,725m; FY2027 $79,581m (high $116,095m, low $57,495m); FY2027 EPS $13.78, +24.2% vs 90 days prior, 8 up / 0 down in 30 days.

Quality, valuation and liquidity can prevent a position on their own. The remaining four inform timing, sizing and monitoring, and never reject an investment by themselves.

Bottom line

The gap between what today's price requires and what the business has demonstrated is -47.1 percentage points, and that gap — not the multiple — is the case that the security is mispriced. The most important unresolved uncertainty is the permanent-loss mechanism: the MI-series accelerator franchise fails to establish a durable merchant share position and the loss is PERMANENT, not cyclical. The next evidence that should change the portfolio decision is the test dated 26 December 2026, or a daily close below $362.59, which forces an immediate review.