Prediction: Data Center Passes 70% of AMD's Revenue in 2027, Before the Helios Ramp Is Finished

AMD's data-center revenue reached $6.7B, or 58% of record $11.5B second-quarter sales, after growing 107% year over year versus roughly 8% growth across its other businesses combined. Third-quarter guidance of about $13B, coupled with management's expectation for accelerating data-center sales, implies the segment could rise to roughly $8.2B and 63% of revenue. Management expects data-center revenue to more than double in 2027 as Helios AI systems ramp for OpenAI, Meta and Anthropic, potentially lifting the segment above 70% of AMD revenue by mid-2027; however, the approximately $474 share price and 30x expected 2027 earnings already assume a smooth Helios rollout.

Analysis

AMD's equity is increasingly a single-product-cycle underwriting exercise rather than a diversified semiconductor story. Rack-scale deployments should raise content per customer and create follow-on demand for HBM, networking, power and liquid cooling, benefiting SK Hynix/Micron, AVGO, ANET and VRT more directly than the headline GPU supplier in some cases. The trade-off is that systems revenue can carry higher integration, warranty and working-capital demands than standalone silicon; revenue mix expansion will not automatically translate into proportional EPS upside unless gross-margin progression confirms it.

The near-term catalyst is evidence that initial rack deployments convert from qualification shipments into repeatable production volume over the next two quarters. The more important 6-18 month risk is customer concentration: a small number of hyperscalers can defer deliveries if power availability, data-center construction, or model-training ROI disappoints. Capacity announcements are not equivalent to binding, financed purchase orders; investors should monitor deferred revenue, receivables, inventory turns, gross margin and customer prepayments rather than extrapolating announced gigawatts.

Consensus appears to frame AMD as a clean second-source beneficiary of AI capex, but the harder question is whether its software and rack-level reliability can sustain utilization economics against NVDA's incumbent ecosystem. If deployments work, AMD's share gains could force NVDA to defend price or bundle networking more aggressively, pressuring industry gross margins even while unit demand remains strong. Conversely, a broad PC recovery would improve AMD earnings diversification and reduce concentration risk, but could dilute the perceived purity of the AI multiple rather than create incremental upside.

At a growth-stock valuation that presumes execution, the favorable setup is conditional rather than an immediate beta chase. A material upside rerating requires quarterly data-center growth to remain above expectations while gross margin expands; revenue growth alone accompanied by flat margins would indicate costly share acquisition. A shipment slip, weaker-than-expected customer capex commentary, or a sequential deterioration in inventory/receivables would falsify the bullish ramp thesis quickly.

Key Decisions for Investors

• Maintain a conditional long AMD position only after the next earnings release confirms sequential data-center growth and at least stable gross margin; target a 15-20% upside rerating on validated production ramps, with a 8-10% stop or exit on a guide-down tied to rack shipments or margin dilution.
• Express AI infrastructure spillover through a basket long VRT and ANET versus short SOXX in a 3-6 month horizon; rack-scale buildouts increase power/cooling and network intensity, while the hedge reduces exposure to a broad GPU-demand reset. Exit if hyperscaler capex guidance turns negative or VRT/ANET orders fail to convert into backlog growth.
• Avoid using META capacity commitments alone as a reason to add META. Instead, monitor its next two capex guides and depreciation outlook: accelerating capex without a corresponding improvement in ad monetization or AI-driven engagement would be a 6-12 month margin-risk signal and supports a tactical META underweight versus GOOGL.
• Watch AMD/NVDA relative performance after the first production deployment disclosures. If AMD shows credible utilization and repeat orders while NVDA maintains pricing, favor long AMD / short NVDA in equal dollar size for 6-12 months; do not initiate before independent evidence on software performance, customer acceptance, and gross-margin economics.