Event date: September 11, 2026
Source: StockAnalysis transcript: https://stockanalysis.com/stocks/amd/transcripts/738878-goldman-sachs-communacopia-technology-conference-2026/
Jim Schneider (Semiconductor Analyst, Goldman Sachs): Good morning, everybody. Welcome to the final day of the Goldman Sachs Communacopia + Technology Conference. It's my pleasure to welcome AMD to the stage. With us from the company, we have SVP and General Manager of Compute and Enterprise AI, Dan McNamara, and Corporate Vice President of Financial Strategy & Investor Relations, Matt Ramsay.
Dan McNamara (SVP and General Manager of Compute and Enterprise AI, AMD): AMD's internal AI adoption is progressing from data-layer optimization and automation agents toward domain-specific applications, including EDA, coding, debugging, kernel development, and software development. AMD is also balancing token costs with productivity gains and applying AI internally to accelerate hardware and software time to market.
Matt Ramsay (Corporate VP of Financial Strategy and Investor Relations, AMD): AMD is using Anthropic Claude tools and other AI tooling across engineering. The company sees faster software development, debugging, and chip-program execution, with productivity gains exceeding the incremental token cost.
Jim Schneider (Semiconductor Analyst, Goldman Sachs): Asked how model evolution may develop across frontier, open-weight, and small language models.
Matt Ramsay (Corporate VP of Financial Strategy and Investor Relations, AMD): AMD is designing its CPU and GPU roadmap for tokens-per-dollar outcomes across open-weight, frontier, and other models.
Dan McNamara (SVP and General Manager of Compute and Enterprise AI, AMD): Enterprises are likely to use hybrid architectures: frontier models in the cloud, open-weight models on premises, and policy-based routing based on performance, latency, cost, and security.
Jim Schneider (Semiconductor Analyst, Goldman Sachs): Asked where AMD can make the biggest strides in AI data-center products, software, customers, and ecosystem.
Dan McNamara (SVP and General Manager of Compute and Enterprise AI, AMD): AMD is shifting from individual product lines to full rack-scale, system-level solutions. A combined data-center roadmap now coordinates CPUs, GPUs, networking, and software around customer workloads.
Matt Ramsay (Corporate VP of Financial Strategy and Investor Relations, AMD): AMD can increasingly act as a strategic high-performance-computing partner, combining CPU, GPU, FPGA, semi-custom, endpoint, cloud, and on-prem capabilities according to customer requirements.
Jim Schneider (Semiconductor Analyst, Goldman Sachs): Asked about the expected diversity of AI customers and near-term markers for 2027.
Matt Ramsay (Corporate VP of Financial Strategy and Investor Relations, AMD): AMD reiterated more than 60% data-center growth and more than 80% AI-business growth targets, said the data-center TAM had expanded from $60B to more than $220B, and said the total company TAM is now around $2T by 2030. Management expects 2027 data-center revenue to more than double and said long-term earnings potential is now significantly above $20 per share.
Dan McNamara (SVP and General Manager of Compute and Enterprise AI, AMD): AI customer concentration is currently high as Helios ramps, but enterprise adoption should broaden over time. Server CPU growth is already broad across cloud, enterprise, and channel customers.
Jim Schneider (Semiconductor Analyst, Goldman Sachs): Asked why agentic AI creates incremental CPU demand and higher CPU tag rates.
Dan McNamara (SVP and General Manager of Compute and Enterprise AI, AMD): Agentic systems add CPU-intensive control-plane, API, database, and tool-execution workloads alongside GPU workloads. AMD sees three CPU lanes: high-core-count agentic sandbox CPUs, high-frequency head-node CPUs, and general-purpose server CPUs. Venice is designed to address these workloads.
Matt Ramsay (Corporate VP of Financial Strategy and Investor Relations, AMD): The server TAM was raised from $60B to $120B and then $220B by 2030, with AMD targeting more than 50% of that opportunity through its product breadth.
Jim Schneider (Semiconductor Analyst, Goldman Sachs): Asked about x86 durability versus ARM and hyperscaler custom silicon.
Dan McNamara (SVP and General Manager of Compute and Enterprise AI, AMD): The issue is optimization rather than ISA alone. AMD sees strong design-ins across cloud vendors, with x86 ecosystem depth and performance-per-watt/per-dollar advantages. Amazon's RDS service was cited as running on Turin where performance is the deciding factor.
Jim Schneider (Semiconductor Analyst, Goldman Sachs): Asked how investors should think about CPU content per gigawatt.
Matt Ramsay (Corporate VP of Financial Strategy and Investor Relations, AMD): AMD is focused on becoming a major partner for inference and agentic workloads regardless of which accelerator runs the model; it did not provide a fixed CPU-to-GPU ratio.
Dan McNamara (SVP and General Manager of Compute and Enterprise AI, AMD): CPU content is workload-dependent. The company sees roughly a 1:1 ratio today in some deployments, with CPU demand growing as agentic workloads expand.
Jim Schneider (Semiconductor Analyst, Goldman Sachs): Asked what investors may find most surprising about AMD in five years.
Dan McNamara (SVP and General Manager of Compute and Enterprise AI, AMD): AMD is becoming more of a software and systems company, with rack-scale offerings and a software stack focused on customer time-to-value.
Matt Ramsay (Corporate VP of Financial Strategy and Investor Relations, AMD): The company is at an inflection point, expecting data-center revenue to more than double next year while gross-margin dollars grow faster than expenses; management emphasized execution over promotional claims.
Jim Schneider (Semiconductor Analyst, Goldman Sachs): Thanked AMD for participating.
Dan McNamara (SVP and General Manager of Compute and Enterprise AI, AMD): Thanked the host.
Matt Ramsay (Corporate VP of Financial Strategy and Investor Relations, AMD): Thanked the host and concluded the session.
*Note: This is a cleaned speaker-labelled rendering of the public StockAnalysis transcript. It preserves the substantive presentation and Q&A but normalizes repeated page-extraction artifacts and spacing.*
End of transcript.
Transcript source limitation: The source page is a conference transcript rather than an earnings call; no quarterly Results/Guidance/Capex block was presented at this event.
Operator / closing: The session concluded after the host thanked the AMD representatives.