logo
|
Blog
  • Yetter
  • OwLite
  • Fits on Chips
  • SqueezeBits
  • 🌐
Event

Our Experience Running a Booth at GTC 2026

Sharing GTC 2026 insights, which is the Largest AI Industry Conference for developers! If you’ve ever wondered what it’s like for an AI startup to run a booth at such a massive event, you won’t want to miss this!
Goeun Kang's avatar
Goeun Kang
Mar 27, 2026
Our Experience Running a Booth at GTC 2026
Contents
Jensen Huang's Keynote and the Rise of Physical AIThe Fastest Diffusion Engine, YetterRoBoost Workflow Agent: Bringing Physical AI to ProductionLessons from Running a Booth at GTCWrapping Up

NVIDIA GTC 2026, one of the AI industry's most important annual events, took place in sunny San Jose, California, this March. The weather was warm, but the energy from developers and companies gathered worldwide felt even warmer.

Now GTC goes beyond a typical conference. It offers the earliest look at the future of AI and accelerated computing. The technologies and directions unveiled here often shape global industry trends, drawing worldwide attention every year.

SqueezeBits has attended GTC every year to gather insights, but this time was different. Instead of just attending, we ran our own booth and presented our technology firsthand.

As an NVIDIA Inception partner, SqueezeBits has built ongoing collaboration within the NVIDIA ecosystem. That work includes developing our open-source framework Ditto. For this GTC, six team members, including our CEO, traveled to San Jose to demonstrate optimization technologies that maximize GPU utilization.

SqueezeBits team members posing at the main hall entrance before GTC kicked off

Jensen Huang's Keynote and the Rise of Physical AI

GTC opened with the keynote from NVIDIA CEO Jensen Huang. Our team arrived early to catch the live session, which let us grab seats near the front.

The standout message was clear: computing is being restructured at the full-stack level, spanning hardware, software, and robotics rather than isolated domains.

Physical AI sat at the center of that vision. AI is expanding beyond text and image generation toward understanding and interacting with the real world. We saw this trend across robotics and autonomous driving as well as simulation. RoBoost, which we debuted at this GTC, felt especially relevant. It connects directly to Physical AI commercialization, going well beyond a simple feature showcase.

A photo from the front row at Jensen Huang's keynote session
NVIDIA CEO Jensen Huang delivering the keynote

The Fastest Diffusion Engine, Yetter

SqueezeBits centered our GTC demo around Yetter, our next-generation inference engine.

With deep strengths in diffusion model optimization, we emphasized that Yetter goes far beyond raw speed. It is a GPU-native unified optimization stack that folds model compression, serving optimization, and hardware-level tuning into one pipeline. This maximizes GPU utilization while cutting memory overhead and boosting inference speed.

The most common reaction during our live booth demos was "This is much faster than I expected." Image generation speed was noticeably fast, and visitors who tried it firsthand walked away impressed.

Yetter supports not only image and video models but also large language models (LLMs) and world models. Many visitors scanned the QR code to try it themselves.

A key highlight was our live demo of a Flux model built on Mojo, a high-performance language co-developed with Modular. Modular is a US unicorn startup with over $300 million in cumulative funding. High-quality image generation took under one second. Booth visitors were struck by the speed, nodding at the "fastest diffusion engine" title. Many scanned the QR code multiple times to keep experimenting.

The SqueezeBits booth at GTC 2026

RoBoost Workflow Agent: Bringing Physical AI to Production

While Yetter demonstrated peak speed, RoBoost made its debut at this GTC and demonstrated AI's scalability.

RoBoost is an agentic workflow developed by the SqueezeBits team. It uses NVIDIA Cosmos-based world models to generate scenario data for training. RoBoost extends Cosmos's image data into video format, producing high-quality scenario data that maintains physical consistency and task-label alignment.

Physical AI systems like robots and autonomous vehicles require massive training data from virtual environments. Collecting diverse real-world scenarios is limited by cost and safety constraints.

RoBoost solves this by generating scenario data that reflects varying conditions such as weather, lighting, and materials while maintaining natural temporal and spatial continuity. This enables more efficient collection of training data that closely mirrors real-world environments.

The on-site response was impressive. Many visitors asked whether the generated data could serve as actual training data, not just synthetic video. Others nodded along while watching the demo videos. Several visitors looked up our technical blog or left their contact information for follow-up consultations. We could feel the demand for Physical AI solutions firsthand.

The SqueezeBits team staying busy as visitors filled the booth
Visitors scanning QR codes to try the Yetter demo and visit our homepage

Lessons from Running a Booth at GTC

After this week, we understood why our engineers who attended last year's GTC strongly recommended it. With developers and researchers from around the world in attendance, the conversations went beyond surface-level interest into specific, in-depth technical questions. More than the visitor count, what stood out was meeting people who genuinely needed our solutions.

Battling jet lag while standing at the booth all day was no small feat. But as we stood together in our signature orange uniforms, we were fueled by a deep sense of camaraderie and pride. For a marketer, there’s truly no substitute for hearing a customer’s voice in person. It’s an experience no office monitor can ever replicate!

The booth reinforced a vital lesson: it’s about far more than just showcasing a product.

Being on the floor all day was an invaluable feedback loop. We learned in real-time which value propositions truly landed and which features sparked the most curiosity. Beyond our own booth, we scouted branding strategies and identified strategic partners for future collaboration, gaining insights that are simply impossible to gather from behind a desk.

Wrapping Up

The five-day schedule was short but fully packed. Despite the jet lag and physical demands, completing the event together as a team made it all worthwhile.

Above all, seeing customer reactions firsthand and explaining our technology opened up new possibilities. That was a marketer's most rewarding experience. We also spent late nights after each day analyzing the companies that visited our booth, sharing feedback, and refining our approach.

GTC 2026 confirmed that SqueezeBits holds globally competitive technology. The on-site team and the colleagues in Korea who handled logistics and prep behind the scenes made the outcome possible together. That shared drive, with everyone pushing to represent SqueezeBits at our absolute strongest, was the true highlight of the week.

We expect the connections and insights from this event to fuel the next phase of SqueezeBits' growth.

Follow us on LinkedIn and our blog for updates on upcoming conferences!

Celebrating a successful GTC 2026 with the SqueezeBits team shining in orange
Share article

The official SqueezeBits Tech blog

RSS·Powered by Inblog