Watch: Nvidia Unveils 'Project GROOT' Robot & New 'Grace Blackwell' Superchip
Nvidia announces the highly anticipated Blackwell platform to power a new era of computing, specifically introducing Blackwell B200 AI chip.
Among the many organizations expected to adopt Blackwell are Amazon Web Services, Dell Technologies, Google, Meta, Microsoft, OpenAI. Oracle. Tesla and xAI.
A Massive Superchip
"The NVIDIA GB200 Grace Blackwell Superchip connects two NVIDIA B200 Tensor Core GPUs to the NVIDIA Grace CPU over a 900GB/S ultra-low-power NVLink chip-to-chip interconnect."
"For the highest AI performance. GB200-powered systems can be connected with the NVIDIA Quantum-X800 InfiniBand and Spectrum™-X800 Ethernet platforms, also announced today, which deliver advanced networking at speeds up to 800Gb/s."
"The GB200 is a key component of the NVIDIA GB200 NVL72. a multi-node, liquid- cooled, rack-scale system for the most compute-intensive workloads."
"The GB200 NVL72 provides up to a 30x performance increase compared to the same number of NVIDIA H100 Tensor Core GPUs for LLM inference workloads, and reduces cost and energy consumption by up to 25x."
B200 AI chip features 208bln transistors.
Just one thing...
B200 recommended power supply pic.twitter.com/AoNGlPcOk1
— zerohedge (@zerohedge) March 18, 2024
All those who spent tens of billions on A100 and H100s capex are now looking at tens of billions in obsolete inventory.
Welcome to the AI cadence cycle.
But there's more - towards the end of the keynote, Jensen Huang announced Project GROOT, which stands for "Generalist Robot 00 Technology".
Huang revealed that robots powered by the platform will be designed to understand natural language and emulate movements by observing human actions, allowing them to quickly learn coordination, dexterity, and other skills to navigate, adapt, and interact with the real world – and absolutely won’t lead to a robot uprising.
Huang then showed a number of demos in which Project GROOT-powered robots performed a number of tasks, showing off its capabilities.
“Building basic models for general humanoid robots is one of the most exciting problems we can solve in AI today,” says Huang.
“The enabling technologies are coming together to allow leading roboticists around the world to make giant leaps toward artificial general purpose robotics.”
The initiative includes the unveiling of Jetson Thor, a computer for humanoid robots based on the NVIDIA Thor SoC, and major upgrades to the NVIDIA Isaac robotics platform.
* * *
As we wrote in our weekly preview note earlier, while this week is chock-full of central bank announcements, including a historic rate hike by the BOJ which will also be ending its YCC and purchases of ETFs, all eyes are on Nvidia which has two key events this week, the first of which may be even more market-moving than the central bank announcements. We are talking about NVIDIA hosting the world's largest and most important conference centered around AI, its GTC (Graphics Technology Conference), where CEO Jensen Huang will talk about the next innovations in the segment and unveil the next generation of GPUs powering the latest cloud and data center markets.
For those unfamiliar, NVIDIA is hosting its latest episode of GTC today at 4pm (right at the market close) and what makes 2024's GTC special is the fact that the company has seen huge successes from the AI segment when it comes to its financial performance. Starting with its Volta V100 GPU family up to the latest Ampere A100 and Hopper H100 chips, the company is the official king of AI.
But at the same time, the market is heating up with competition hitting back on both the hardware and software sides. Keeping that in mind, this GTC, NVIDIA is not only going to respond to its competition with new hardware innovations but will also certainly showcase how its AI software strategy has helped define its leadership in this segment and how it will evolve over the coming years.
According to wccftech, some of the new products that we can expect to see at the keynote will be the Hopper H200 GPUs featuring the fastest HBM3e memory with upgraded capacities, and also the first showcase of the truly next-gen Blackwell B100/B200 GPUs. There are reports that the first generation of Blackwell GPUs will feature HBM3e memory with huge computing and AI-side upgrades while the B200 GPUs will take things to the next level in the coming year with updated HBM4 specs and even more software optimizations.
For those who want to tune into the event, the GTC 2024 keynote will be live-streamed on 18th March (Monday) at 1 PM (Pacific Daylight Time) and last 2 hours till 3 PM so one can expect a very jam-packed presentation by NVIDIA CEO, Jensen Huang. In addition to being live-streamed, this will be the first GTC after many years that will have a proper in-person attendance and we can expect lots of key executives of other companies to show up to announce their support and partnership with NVIDIA.
* * *
In any event we are not the only ones who view today's GTC presentation as the biggest market event of the week: according to Goldman trader Cullen Morgan, the NVDA GTC Conference presentation by CEO Jensen Huang on Monday 4pm ET will be the most important macro event of the week, with NVDA options implying a +/- 9% move over the next 1-week; which suggests investors expect the presentation to be as impactful as a typical earnings event.
The chart below shows that 1-week NVDA implied volatility is higher than ahead of two of the past four earnings events. At the index level, NDX 1-week options are pricing an +/- 2.2% move over the next week and a +/- 1.5% move between now and Tuesday’s close. As Morgan concludes "the Tuesday morning reaction to the NVDA presentation and the Wednesday afternoon reaction to the FOMC statement will likely be the two most important hours of the week for the market."
Meanwhile, according to UBS trader Robert Ruple, consistent with the feedback from his recent Nvidia meeting, focal points will likely center around:
i) the growth of Inference (stated that around 70% of current total Data Center GPU mix) including the adoption of AI across end-markets with an expanding total addressable market (including greater Enterprise adoption),
ii) updates on their roadmap (B100/N100),
iii) monetization opportunities from their CUDA/software ecosystem, while addressing the competitive backdrop and China.
Structurally, Ruple remains bullish on Nvidia, but he does question if this could be a tactical “sell-the-news” event given all the hype with the stock up 77% year to date and 243% over the last year. Still, it’s tough to fight “city hall” nor the overwhelming enthusiasm for Nvidia/other AI stocks these days.
Needless to say, traders are on edge, and in a slightly scaled down version of the recent super wipeout in NVDA stock which saw the company lose a quarter trillion in market cap in one day on March 8, today we have seen a gain of as much as 5.2% evaporate as NVDA stock turned from bright green to red, wiping out tens of billions in market cap, as traders prepare for possible downside after Huang's appearance later today.
This would be too easy, right?
We will be live streaming the speech when it starts at 4pm ET.