What did Nvidia say at the Berlin GTC? Here’s Goldman Sachs’ summary

On October 21, Nvidia CEO Jensen Huang delivered a keynote speech at the GTC Berlin conference, after which the Goldman Sachs research team released a report.

The report points out that this speech is expected to have a limited immediate impact on stock prices, but Nvidia’s core message “reinforces our view—AI deployment remains on track for the foreseeable future.”

Key announcements from the conference include: Nvidia launched the DSX POD modular data center solution, compressing deployment cycles from several months to two weeks; supply chain partners announced a 2.5x increase in substrate capacity and a 3x expansion in CoWoS packaging; physical AI business revenue for the first half of FY27 has reached $11 billion, surpassing the full fiscal year FY26; and the release of the OpenShell architecture addressing enterprise AI security issues.

Goldman Sachs maintains a “Buy” rating on Nvidia with a 12-month price target of $300, representing about 38.6% upside from the current price of $216.47. The report also emphasizes that Nvidia continues to provide strong catalysts for its stock price by narrowing the supply-demand gap and driving robust capital returns (with approximately $235 billion in buyback authorization).

Clients Are Earning Profits, AI Deployment Shows No Sign of Slowing Down

The report notes that Nvidia’s customers—including hyperscalers and neoclouds—“are achieving strong unit economics from current data center deployments.”

In other words, the money spent is being recouped. This is the core logic behind the sustainability of AI capital expenditure.

Goldman Sachs therefore judges that AI deployment “remains on track for the foreseeable future.”

Massive Supply Chain Expansion, Preparing for Larger-Scale GPU Shipments

At the conference, Nvidia announced joint efforts with several partners to expand supply chain capacity across key steps. Specifically:

• Substrate suppliers Ibiden, SEMCO, and Unimicron will increase total capacity by 2.5 times;

• Amkor’s CoWoS-R packaging capacity will be increased 3-fold;

• KYEC will boost US-based capacity;

• SPIL will significantly expand CPO (co-packaged optics) capacity.

Analysts believe there are deeper signals behind this expansion. The report writes, “This suggests Nvidia is preparing for a large-scale CPU shipment far beyond prior expectations.” In other words, the scale of this supply chain expansion exceeds general market expectations for the ramp-up of Nvidia’s next-generation products.

DSX POD: Cutting Data Center Deployment Time from Months to Two Weeks

Nvidia and partners introduced the DSX POD reference design—a containerized module capable of accommodating multiple racks of the complete Vera Rubin system, fully factory-assembled and ready to run on-site upon power-up.

This product is expected to hit the market in early 2027. Analysts point out that the core value of DSX POD lies in slashing data center deployment time from “months” to “two weeks”, significantly shortening the period from deployment to initial token output.

Efficiency improvements are equally noteworthy. Nvidia disclosed that its DSX Factory Platform has enabled operators (including neoclouds) to achieve about 30% power savings in deployments totaling 2GW, as well as daily water savings of 56 million liters.

AI Security Challenges—Nvidia Offers a Hardware-Level Solution

As AI agents penetrate enterprises at an accelerated pace, security issues are becoming increasingly prominent. Nvidia announced a series of new software releases at the conference, most notably the OpenShell reference architecture.

Analysts describe that the core logic of the platform is to “keep the granting of critical permissions outside the agent itself, enforcing policies via hardware triggers to prevent any unauthorized use.” This means that even if the AI agent is compromised or misused, the hardware-level restrictions still function as a final line of defense.

OpenShell integrates Vera Rubin, Vera CPU, and BlueField-4 DPU (with Sentry module) to form a comprehensive security loop from chip to software. Goldman Sachs believes this arrangement helps Nvidia build a deeper moat in the enterprise AI market.

Physical AI Revenue Surges—First Half FY27 Exceeds Last Full Year by 30%

Physical AI was another major theme at this conference for Nvidia. The company announced several initiatives to accelerate physical AI adoption, including the Hyperion reference design for autonomous vehicles and Robotaxis, the AGILE ONE reference design for humanoid robots, and AI-based telecom infrastructure platforms.

In terms of figures, Nvidia revealed that first-half FY27 physical AI revenue reached $11 billion, approximately 30% higher than all of FY26.

Analysts see this data as powerful evidence of Nvidia’s diversified revenue structure. Outside data center operations, physical AI is fast becoming a second engine of growth for Nvidia.

Goldman Sachs Maintains Buy Rating, $300 Target Price, Nearly 40% Upside

Goldman Sachs maintains its Buy rating on Nvidia, with a 12-month price target of $300 based on a 30x P/E ratio multiplied by a normalized EPS estimate of $10.

From a forecast perspective, Goldman Sachs expects Nvidia’s FY27E (ending January 2027) revenue to reach $412.7 billion with EPS of $9.40; FY28E revenue is projected to rise further to $718.3 billion with EPS at $16.70. The current share price equates to an FY27E P/E ratio of about 23x and FY28E of about 13x.

Analysts also list major downside risks including: slower AI infrastructure spending, intensifying competition leading to market share loss, margin compression under competitive pressure, and supply chain constraints.

Nvidia’s current market cap is about $5.3 trillion, with approximately $235 billion in authorized stock buyback plans.