AMD's Last Three Years: From Chasing Numbers to Naming Partners

The story

Late 2023 into 2024: a company that spoke in specific numbers

At the end of 2023, AMD spoke the language of specific numbers. In its January 2024 earnings call, it stated flatly that "2024 AI accelerator (data center GPU) revenue will exceed $3.5 billion" — then raised that figure to $4 billion in April, $4.5 billion in July, and $5 billion in October, raising the target every single quarter. That year's 10-K still opened its "Our Strategy" section with the somewhat bureaucratic phrase "five strategic pillars," and risk factor #1 was "Intel's dominance of the microprocessor market" — AMD was still describing itself as the challenger fighting in Intel's shadow.

February 2025: an analyst asks directly, and the numbers go away

On the February 2025 earnings call, analyst Vivek Arya asked directly: "Why no specific target this year — what's changed?" CEO Lisa Su answered that guidance would move to a segment-level basis going forward, and the number that had been raised four consecutive quarters quietly disappeared. Two months later, in April 2025, new U.S. government restrictions on AI chip exports to China forced AMD to absorb roughly $800 million in inventory losses on its MI308 product, and words like "headwind" and "dynamic" started appearing noticeably more often in earnings-call language.

AMD's response: rewrite the story, not re-issue the number

Instead of resurrecting specific targets, AMD rewrote its own narrative. Starting with the FY2024 10-K, the opening line of "Our Strategy" was completely rewritten to "AI is defining the next era of computing," and that same year's risk factors added "Nvidia's dominance of the GPU market" to the top of the list for the first time. But the FY2025 10-K went further still — both the "Intel's dominance" and "Nvidia's dominance" language were removed entirely, replaced with a blander "intensely competitive and rapidly evolving markets" framing. Filling that vacated space instead: an October 2025 agreement with OpenAI for "6 gigawatts of GPU supply, with over $100 billion in potential revenue over the coming years," followed by partnerships with Meta (Q1 FY2026) and Anthropic (Q2 FY2026).

Results actually got stronger throughout

Through all of this, results kept improving. Of 11 tracked quarterly guidance cycles since late 2023, not one has been missed, and since the OpenAI announcement (Q3 FY2025 onward), AMD has beaten the top end of its guidance by $200-400 million for three straight quarters. Management's qualitative confidence score dropped to around 7 in early 2025 before climbing back to 9 by 2026, with 2027 Data Center segment growth now described as likely to run "well above 100%."

Guidance scorecard

11 for 11 — no misses at all, and since the OpenAI deal was announced, the margin of the beats has widened noticeably. This is the classic conservative-guide-then-beat pattern, but the growing size of the beats suggests demand may be outpacing even AMD's own internal forecasts.

Source: earnings call guidance statements and the following quarter's reported results, Q3 FY2023 through Q2 FY2026.

Timeline

• Jan–Oct 20242024 AI-accelerator revenue target raised four consecutive quarters, from $3.5B to $5B — backed by results each time.
• Feb 2025Specific annual AI-GPU target discontinued after an analyst asks why directly.
• Apr 2025New U.S.-China export controls force roughly $800M in MI308 inventory losses.
• Oct 2025OpenAI partnership announced: 6GW GPU supply, $100B+ potential revenue over multiple years.
• 2026Meta partnership (Q1) and Anthropic partnership (Q2) added; guidance beats widen; management confidence score reaches 9.

Our read

AMD's last three years trace a shift from proving itself through specific, repeatedly-raised numbers to proving itself through named, marquee partnerships — a move that coincided with both companies it once named as threats (Intel and Nvidia) disappearing from its own risk-factor language. The guidance record backs the shift up: not a single miss in 11 quarters, with the margin of overshoot growing since the OpenAI deal. The flip side is concentration — the story now rests heavily on whether a handful of enormous, multi-year contracts actually convert into revenue at the pace described.

What we still don't know

• How much revenue, and on what timeline, the OpenAI, Meta, and Anthropic partnerships will actually generate requires watching results from Q3 FY2026 onward — everything so far is described as "planned" or "early shipments."
• Why AMD stopped giving specific targets exactly when it did was never explained in detail publicly — whether demand became harder to forecast or the company wanted to avoid tipping off competitors is speculation either way.
• Whether removing "Nvidia's dominance" language from the risk factors reflects an actual narrowing of the competitive gap can't be judged from this analysis alone — it would require comparing against Nvidia's own filings and results.