Compute & Infrastructure
Designs data-center networking, storage, and custom AI silicon that move and process data inside cloud infrastructure.
{'verdict': '2 signals sit in the elevated band: the AI-monetization gap and organic-demand sustainability. This does not trip the convergence flag. Ranks 21st of 68 on composite fragility (F\xa051), below SoundHound AI and MongoDB.', 'as_of': '2026-07-11', 'source': 'engine-restatement (T1)', 'snapshot': {'composite_f': 51.0, 'n_elevated': 2, 'convergence': 'moderate', 'rank': 21, 'elevated': ['the AI-monetization gap', 'organic-demand sustainability']}}
Top-10 customers are 81% of FY2025 revenue and a single distributor is 34%. If one hyperscaler pauses its custom silicon program or insources — a risk your own 10-K names explicitly — what does that do to revenue?
The concentration is confirmed primary: top-10 customers 81%, Distributor A 34% (up from 24% FY2024, 20% FY2023), from the FY2025 10-K customer concentration note. The 10-K explicitly names as a risk that customers may "develop their own solutions, vertically integrate which may reduce the need for our products." Non-DC segments that could buffer a pause have already collapsed: carrier revenue fell 68% and enterprise networking fell 49% in FY2025. There is no meaningful revenue cushion outside the hyperscaler cohort.
You say all four U.S. hyperscalers are in production for custom silicon, but Google-specific revenue is labeled NOT SOURCED in this sheet. How is that diversification when the filing does not disclose what Google contributes?
The gap is confirmed in the sheet: per-hyperscaler revenue percentages are NOT DISCLOSED in SEC filings. The AWS five-year multi-generational agreement (Dec 2024) is sourced primary. Google custom-ASIC revenue — or any signed contract dollar amount — is NOT SOURCED; press and analyst speculation exists but cannot be cited as fact on the page. The claim of four-hyperscaler diversification is real; its revenue weight is unverifiable from filings.
Management guided $11B FY2027 and $15B FY2028 on the earnings call — those are call statements, not filed guidance. The stock trades at a forward P/E of 61–77× versus an industry median of ~37×. What happens to the multiple if those numbers slip?
The outer-year targets ($11B FY2027, $15B FY2028) are Q4 FY2026 earnings-call statements — the sheet labels them call guidance, not filed forecasts, and they are not in the 8-K body. FY2026 10-K customer concentration is NOT SOURCED for this pass; only the FY2025 81% figure is audited. The forward P/E range of 61–77× (vs. ~37× industry median per GuruFocus) embeds those unaudited targets. Goodwill is $11.1B and total debt ~$4.5B. A concentration-driven revenue miss compresses the multiple and tests the balance sheet simultaneously.