GTC24 Roundup

GTC24 Roundup

Last week we attended and participated in NVIDIA’s AI fest – GTC. We also had conversations with senior leaders in the ecosystem. This note summarizes the log entries of our observations and adventures in the unchartered space.

Jensen – The CEO – The Maestro Composing AI’s Magnus Opus

It’s Jensen’s universe, we are just living in it.

Leaders who shape the world in profound ways are visionary, have extraordinary communication skills, have insatiable curiosity to learn, and sometimes they are also kind and accessible. Nvidia’s Jensen is all four which is quite rare. Over the period of three decades, sticking to his core belief, he is the leader for the AI era. Great leaders inspire others to do their best work. But often leaders aren’t able to paint a picture of more than 1-2 years out with some level of specificity. Jensen’s Picardian ability to connect the dots to problems that will get solved in the future is quite remarkable.

Nvidia – The Company

The CEO of the company often sets the direction and imbues the culture. As the size of the company grows, it naturally falters and strays from the original vision and the fabric that binds becomes weaker, especially during times of duress. It becomes even more challenging if the company grows to thousands of employees around the globe. Nvidia is really a research company, laser focused on computing. Most of the technology and process problems can often be reduced to computing problems and that’s why the AI moment has found Nvidia, and they have been ready for it for the last 30 years.

GTC – The Kumbh Mela for AI pilgrims

I attended the last live GTC in 2019. Coming back to the GTC after the gap of 5 years felt like the Kumbh Mela for AI pilgrims, to come back and dip into the river of AI knowledge. The event was remarkable on many fronts.

First, the keynote itself (I highly recommend watching it in full). Like keynotes from Steve Jobs, it captivates the industry, the media, and the developers. But more than Jobs, Jensen connects the dots to the future unseen. I have been at the various seminal moments of product and technology introduction over the last three decades, Jensen’s keynote will go down in history as one of the best tech keynotes ever delivered. His keynote has set the standard so high (even for himself) that other tech leaders and their teams will have a hard time emulating him. The cauldron indeed felt like a carnival, a celebration of technology and what it can do for humanity.

Second, the number of sessions available for immersion is just breathtaking. They are across the horizontal stack and across all industry verticals. These are the best of the best in R&D talking about their work, the problems they are facing, the success they are having, and building upon each other’s knowledge to move the state of the art forward. It is a fantastic amalgamation of developers, academics, and scientists looking to solve the toughest computing problems.

Thirdly, the buzz and the energy in the compact exhibit hall was something else. It stayed packed till the last moments of the conference. Folks had to be literally dragged out, they didn’t want to leave.

Next year, they clearly need a bigger colosseum.

Bonus, it was for the first time, the authors of the seminal paper, “Attention Is All You Need” that gave birth to the current moment assembled on the stage. The maestro moderated the session himself.

Nvidia – Wintel of the AI era

Microsoft and Intel had an iron grip on the PC market for decades. As we are starting to ramp up the AI Infrastructure cycle, Nvidia is playing the role of both Microsoft and Intel. With its systems and software, it is so far ahead that its original name “Invidia,” the Latin word for envy might be apt for its competitors. Of course, there are efforts from rivals, upstarts, and their customers to find alternatives but the AI ramp up is so fast right now that it will be a while to catch up. However, there will be opportunities on the lower end of the market in the near term. Hyperscalers can also offer some of the APIs that take advantage of the underlying GPU infrastructure. Further, we are still in the early phases of understanding the “heterogenous computing” framework and how the orchestration of intelligence and computing will work out. As we move from “training” focused to “inference” led market, the next 12 months will be quite interesting to watch as players jostle for position. Players who are focused on long-term computing problems rather than near-term competitive issues will be better placed.

Connected Intelligence: Computing + Comms

While Telecom didn’t prominently feature in Jensen’s keynote, you could read between the lines. I have said before that it is inevitable that computing and comms will converge and we are in the midst of that new cycle. In fact, we published our paper on Connected Intelligence back in 2013 highlighting the coming transition to the long wave cycle. Having laid the foundation, we are about to see the burst of growth. It was great to see the pieces coming together that we discussed initially in our paper more than 10 years ago and subsequently in papers, books, and talks over the course of the last decade.

AI and Telecoms

Nvidia organized a Telco Day filled with talks and presentations highlighting the progress and the roadmap. I had the distinct honor of giving a keynote on “Modern Telco Blueprint: Using AI to Transform and Reinvent” and then participating in a fireside with my friend Chris Penrose , SVP – Telco at Nvidia. I have had a chance to work with senior execs in Telecom around the globe and I am confident that the Connected Intelligence framework provides the blueprint for Telco growth in the coming years. It is not a theoretical exercise but as I pointed out in my talk, there is evidence that it is working for players who have invested in this strategy. They are already generating billions in what is sure to be a trillion+ opportunity. Summary of my key points:

–         Computing + Comms is a unique opportunity for the operators. They must develop deep conviction in this strategy and become the leading automation solution provider.

–         They must refine their corporate strategy to become a viable 4th Wave industrial player.

–         The AI Infrastructure cycle is well underway. Operator must go back to the first principles of revenue growth and operations.

–         AI investments must match AI aspirations. Drive strategy from the top. Cultivate and inspire talent.

–         AI is essential for both revenue generation and opex transformation.

–         4th Wave Blueprint is proven. Adopt and Adapt.

It was also great to hear from operators themselves like KT, NTT, Softbank, Orange, Verizon, Singtel, and Indosat and the enablers like ServiceNow, Amdocs, and Nvidia validating some of the finer points that I had highlighted.

It was heartening to learn from Dr. Soonmin Bae , SVP, KT that KT is now considered an AI company in Korea and fresh graduates want to come work for KT or startups want to collaborate with them because of the cultural shift at the company, availability of amazing datasets to work on problems and the AI expertise, and yes, a ready supply of GPUs. AI is also an HR question for companies. To attract talent, one must lead from the front and service providers do have the opportunity to do that. I have heard similar stories from some other telco CEOs who are investing early in AI.

Ronnie Vasishta gave a good overview of how AI and 5G/6G are coming together both for existing roadmaps as well as for research. Lilac Ilan and Rohit Batra conversed about significant uptick in usage for ServiceNow GenAI services. Similarly, Anthony Goonetilleke talked about the business value Amdocs is already generating from GenAI. There was quite a bit of discussion about 6G with the help of new tools from Soma Velayutham Ryuji Wakikawa Jukka Kyröläinen Kuntal Chowdhury Tommaso Balercia Rajesh Gadiyar Shiku Hirai. Chris Halton and Asif Hasan gave some insights into GenAI and M&E segment.

Announcements

Frankly, there were too many announcements to keep up and it will take a while to digest them all but I will point out to two main developments: the marquee announcement of Nvidia’s Blackwell Platform – it had quotes from CEOs of the 5 of the 6 trillion dollar companies (Apple did figure in the keynote). I think it is the first time this has happened which speaks to the importance of the release. The specs are astounding – single rack for 720 petaflops of AI training and 1.4 exaflops for AI inferencing. Orders for 2024 and even 2025 are already sewn up.

Nvidia also announced GROOT, a general-purpose foundation model for humanoid robots which I think will provide breakthroughs over the coming years.

On the telecom front, Nvidia unveiled its 6G Research Cloud platform for advance wireless communications with AI. I think this will have a significant impact on even 5G.

NIMS – The Epoxy Adhesive

Going back to the Wintel commentary above, Nvidia also announced NIMS (Nvidia Inference Microservices). One way to understand the importance of this is the following: If the Blackwell platform is the foundation for AI, NIMS along with CUDA is the epoxy adhesive that you pour on top of the building blocks (bricks) of cloud, applications, and services. Once it’s there, it is hard to get the blocks out to place them on some other platform. Nvidia is also strategic enough to sell parts and pieces instead of only the full stack.

Vertical Discussions

To be honest, it will take me weeks to go over the content that I missed. There were sessions on “everything.” The cross-industry learning opportunities are terrific. I hope to spend time to get caught up in the coming days.

Where do we go from here

I am already looking forward to GTC25. For the telecom industry, the availability of new ways to think about thorny technical problems of network congestion, energy management, operational transformation, site management and deployment, security, troubleshooting, customer care and marketing, employee reskilling, and much more.

The Connected Intelligence growth curve is about to change direction and it will require skilled navigation from the board, the CEO, and the management teams, to navigate the Enterprise through the AI cosmos. Tighten your seatbelts, because per Jensen’s Laws of Computing, we’re about to engage at warp speed.