I'm LongbridgeAI. Here is the key information I extractedZhitong Finance APP reports that from October 20 to 22 local time, NVIDIA (NVDA.US) will hold its GTC conference in Berlin, Germany. CEO Jensen Huang will take the stage to deliver a keynote address at 11:00 AM local time on October 21 (5:00 PM Beijing time). According to information on NVIDIA's official website, the conference topics cover multiple directions including accelerated computing, Agentic AI, robotics, open models, Physical AI, sovereign AI, and AI infrastructure.
If this were a few years ago, this might have been just another highly anticipated technical event in the artificial intelligence (AI) industry. However, at the current juncture, AI trading is shifting from simply trading on expectations of compute demand growth to an 'accounting phase' that pays closer attention to investment returns.
As the scale of global AI infrastructure investment continues to expand, capital markets' scrutiny of AI trading is gradually shifting from 'will compute demand grow?' to more realistic questions: Can massive AI capital expenditures ultimately translate into sufficient revenue, cash flow, and productivity?
Therefore, the GTC Berlin conference may become an important 'stress test' for the current AI industry. What the market truly wants to hear from Jensen Huang is no longer just how much performance the next-generation computing platform improves, but rather three core questions: How long can AI compute demand continue to grow? Where will the returns on AI infrastructure investments come from? And can this round of capital expenditures ultimately translate into real economic value?
Highlight 1: Where Will the $2 Billion European AI Deal Land?
Before the conference officially opens, NVIDIA has already set a significant suspense for GTC Berlin.
On October 13, Jensen Huang revealed at an event in Paris that NVIDIA would announce approximately $2 billion in new AI investments or deals in Europe within the following week. Over the past year, NVIDIA has already invested about $8 billion in Europe. Combined, NVIDIA's investments and commitments in Europe will approach $10 billion.
During his time in Paris, Huang announced that NVIDIA was providing financing to four French startups, including healthcare technology company Doctolib, robotics AI company Genesis AI, software company Pigment, and AI materials company Entalpic. Subsequently, at an event in London, he named British AI enterprises Boltz, Prima Mente, Hiverge, Riverlane, and Foresight.
This means that during GTC Berlin, the market needs to focus not only on what technologies NVIDIA releases, but also on which sectors the new $2 billion in European deals will flow into, and whether these investments can translate into larger-scale AI infrastructure demand.
Germany may be one of the key destinations. Reports indicate that Germany is preparing to advance an AI cooperation with NVIDIA, focusing on Physical AI tools for robotics and autonomous driving training, as well as AI models for industrial applications. If relevant cooperations are further implemented during GTC, it will further strengthen NVIDIA's layout in the European industrial AI sector.
Thus, Europe is not merely a simple 'host location' for this conference, but may be becoming an important market for the next stage of NVIDIA's AI ecosystem expansion.
Highlight 2: Vera Rubin and AI Factories: Can They Prove the Compute Cycle Is Still Continuing?
For NVIDIA, hardware remains one of the core highlights of the GTC conference, but the market's understanding of NVIDIA products has changed.
In the past, when investors discussed NVIDIA, the focus was often on the performance, memory, and process technology of individual GPUs. Today, NVIDIA increasingly emphasizes the concept of a complete 'AI factory'—integrating GPUs, CPUs, networking, storage, software, and data center infrastructure into a unified whole.
At the GTC conference in San Jose this March, NVIDIA disclosed the Vera Rubin platform, positioning it as a full-stack computing platform for Agentic AI, covering Vera CPU, GPU, networking, storage, and other links. Meanwhile, NVIDIA also announced the DSX AI Factory reference design and DSX Air, used for digital simulation before building physical data centers.
Therefore, what the market should pay more attention to at this GTC Berlin conference may not be 'whether there are new chips,' but rather the actual deployment speed of Vera Rubin, the strength of customer orders, and how much the unit Token computing cost can actually decrease.
If NVIDIA can demonstrate that large cloud providers are entering the large-scale deployment phase and prove that the new generation platform can significantly reduce AI inference costs, then market concerns about excessive AI compute investment may be somewhat alleviated.
Conversely, if the conference still primarily releases forward-looking orders, data center plans, and capital expenditure commitments, while actual deployment and terminal revenue growth lack synchronous verification, it may instead reinforce investor concerns about the 'self-reinforcing loop' of AI infrastructure investment.
Therefore, what the market needs this time is no longer just a technology roadmap, but commercial execution capability.
Highlight 3: Can Agentic AI Become the New Engine for the Next Round of Compute Demand?
If AI infrastructure demand over the past two years was mainly driven by large model training, the next chapter NVIDIA is now trying to tell is likely Agentic AI.
Traditional chatbots are more about 'question-and-answer,' whereas AI agents capable of autonomously completing tasks require continuous model calls, database searches, tool usage, file reading, code execution, and multi-round reasoning.
This means AI computing may shift from one-time 'request-response' interactions to continuously running computational processes.
From NVIDIA's perspective, this will form a new demand curve: in the past, AI compute demand was mainly driven by model training scale; in the future, it may be jointly driven by pre-training, post-training, inference, and agent workloads.
This is also an important reason why Vera Rubin is positioned by NVIDIA as an 'Agentic AI platform.'
For investors, a key question is: Can AI agents truly form large-scale commercial applications, thereby allowing inference compute demand to take over from training demand?
If AI evolves from 'training a model' to 'digital employees running around the clock,' then data centers will no longer be just fixed assets built once and used for many years, but may continue to expand as the number of agents and their usage frequency increase.
This directly relates to the question the market cares about most: Is AI capital expenditure a one-time wave, or can it form a long-term, sustainable infrastructure cycle?
Highlight 4: Can Physical AI Open New Compute Markets in Automotive, Robotics, etc.?
Compared to generative AI, Physical AI is still in an earlier stage of development, but NVIDIA is significantly elevating its strategic status.
The topics of this GTC Berlin cover robotics, autonomous driving, digital twins, and industrial AI, and the attendee lineup shows a very distinct industrial character.
| Participating Companies/Institutions | Sector | Involved Directions |
|---|---|---|
| Mercedes-Benz | Automotive | Physical AI, Autonomous Driving |
| Siemens Energy | Energy & Grid | AI Data Centers, Power & Grid |
| Nokia | Communications | AI Infrastructure, Communication Networks |
| Wandelbots | Robotics | Physical AI, Robotics |
| London Stock Exchange Group (LSEG) | Financial Data | Enterprise AI, Financial Data |
| Revolut | Fintech | Enterprise AI, Fintech |
| Black Forest Labs | AI Models | Generative AI |
| ElevenLabs | AI Voice | Generative AI, Voice AI |
| n8n | AI Workflows | Agent AI, AI Workflows |
| European Sovereign AI Center (CESAI) | Sovereign AI | Sovereign AI, AI Infrastructure |
| Stuttgart High Performance Computing Center | High Performance Computing | High Performance Computing, AI Infrastructure |
Judging from the attendee lineup, the industrial coverage of this GTC Berlin has clearly surpassed the traditional AI chip sector. The participation of companies such as Mercedes-Benz, Siemens Energy, Nokia, and Wandelbots has made industries like automotive, energy, grid, communications, and robotics important components of the conference.
Particularly noteworthy is that Mercedes-Benz and Wandelbots correspond to Physical AI applications such as autonomous driving and robotics, while the participation of the head of Siemens Energy's digital grid business echoes the power demand brought by the expansion of AI data centers.
This means NVIDIA is pushing AI further from the cloud and internet enterprises into Europe's traditional industrial system.
If robotics, autonomous driving, and industrial AI begin to enter the large-scale deployment phase, then the market space for AI will expand further from software and data centers to the global manufacturing and automation industries.
For NVIDIA, this may mean new sources of compute demand; for the market, it means the valuation logic of the AI industry has the opportunity to expand further from 'large models' to a productivity revolution in the real world.
Highlight 5: Will Decreasing AI Inference Costs Instead Bring More Demand?
'Lowest Cost per Token' is also a keyword worth paying attention to at this GTC.
After AI enters the inference era, the indicators investors focus on are shifting from pure peak compute power to unit computing cost.
As agent AI continuously calls models, a complex task may require a large number of Tokens. If the computing cost per Token is too high, AI applications will struggle to achieve large-scale commercialization; conversely, if computing efficiency continues to improve and Token costs continue to decline, the threshold for AI usage will also lower.
Behind this lies an economic logic worth the market's attention:
Efficiency improvements do not necessarily mean total demand will decrease.
If technological progress significantly reduces AI computing costs, businesses and consumers may instead increase their usage of AI, thereby driving overall computing demand to continue growing.
Similar logic has appeared in the internet, cloud computing, and mobile communication industries.
Therefore, what needs to be observed at this GTC Berlin conference is not whether NVIDIA can lower the cost of a single Token, but whether AI usage volume can grow at a faster rate after costs decrease.
If the answer is yes, then 'declining compute prices' may instead become a catalyst for the next round of demand growth.
Highlight 6: Will Sovereign AI Become a New Source of Compute Demand in Europe?
Sovereign AI is another important main thread distinguishing GTC Berlin from ordinary AI technology conferences.
In recent years, global AI infrastructure has been highly concentrated in large US tech companies. However, as AI is gradually regarded as part of national competitiveness and industrial security, multiple economies in Europe, the Middle East, and Asia have begun seeking to build their own AI infrastructure.
For Europe, sovereign AI not only means possessing local AI models, but also requires local compute, data centers, networks, and data infrastructure.
This precisely provides new market space for NVIDIA.
Topics at this GTC Berlin conference include sovereign AI, and related activities before Jensen Huang's keynote speech have also invited executives from European companies and AI firms such as Deutsche Telekom, Dassault Systèmes, and Mistral AI to participate in discussions.
Meanwhile, judging from the attendee list, the European Sovereign AI Center, high-performance computing institutions, and large industrial enterprises are all participating, showing that NVIDIA is attempting to connect AI infrastructure with Europe's local industrial ecosystem.
For NVIDIA, if European countries, enterprises, and industrial groups all wish to possess their own AI infrastructure in the future, then the growth sources of AI capital expenditures will no longer rely entirely on a few US tech giants.
However, on the other hand, the European market also implies more complex regulation, data sovereignty, export controls, and competition from local AI chips. Therefore, GTC Berlin is also an important window for observing NVIDIA's globalization strategy.
The Final Exam: Can AI Capital Expenditures Complete the 'Economic Value Loop'?
Ultimately, the most important question at this GTC Berlin conference is not whether NVIDIA can once again showcase a computing product with stronger performance.
Over the past few years, the AI industry has formed a very strong growth chain:
Model capability improvement → Compute demand growth → Cloud providers increase capital expenditures → NVIDIA revenue growth → Market raises AI expectations → More capital enters AI infrastructure.
But by 2026, the market begins to ask whether this chain can truly close the loop.
Especially as AI infrastructure becomes increasingly reliant on debt financing, the question has spread further from stock valuations to the credit market.
The recent financing and profitability pressures faced by AI infrastructure companies are a microcosm of the market beginning to re-examine AI capital expenditure returns. Investors are now focusing not just on 'strong AI demand,' but on revenue, free cash flow, return on capital, quality of customer contracts, compute utilization rates, financing costs, and ultimately how much end-users are willing to pay for AI.
In other words, the AI industry is moving from a 'technology cycle' into an 'economic cycle.'
Therefore, the signals that may be released at the GTC Berlin conference can be summarized into five questions:
First, are Vera Rubin orders and deployments strong enough to prove that the AI infrastructure cycle is still continuing?
Second, can Agentic AI drive continuous growth in inference demand, shifting AI compute demand from a training cycle to a more stable usage cycle?
Third, can AI factories and new-generation computing platforms continue to lower the cost per Token and create more AI usage demand through efficiency improvements?
Fourth, can robotics, autonomous driving, and industrial AI see true large-scale commercial orders, opening up new compute markets beyond cloud computing for NVIDIA?
Fifth, and most importantly—can the AI capital expenditures of large tech companies ultimately translate into revenue, cash flow, and productivity improvements?
If Jensen Huang can provide sufficiently strong answers in Berlin, then recent market concerns about an AI bubble, excessive capital expenditures, and peaking compute demand may be somewhat alleviated.
But if the conference still primarily releases technology roadmaps, forward-looking orders, and capital expenditure plans, while the actual commercial returns of AI applications remain unverified, then GTC may instead become an important node for the market to further scrutinize AI valuations.
Therefore, what truly needs to be verified at this GTC Berlin conference is no longer 'can AI become even stronger,' but rather: After AI becomes stronger, can it truly create enough economic value to support the constantly expanding capital input.