May 14, 2025
10 mins read
Ctrl+Alt+Con: The Week Techies Had to Choose Between GDC and GTC
GDC & GTC: The Conference Mix-Up
In March 2025, two major tech events went head-to-head: NVIDIA’s GTC in San Jose and the Game Developers Conference (GDC) in San Francisco. Just an hour apart but worlds away in tone, style, and crowd.
Picture it: A game developer shows up to what he thinks is GDC, ready to demo his new build, trade Unity tips, and pick up a few pixel art stickers. But as he approaches the venue, something feels ... off. Too many suits. Too many slide decks. Not enough noise. Turns out, he’s wandered into NVIDIA GTC, where the vibe is all AI inference, digital twins, and data center–scale computing.
Across the valley, an AI engineer walks into GDC expecting deep technical sessions on GPU optimization, only to be met with neon graphic tees and a roundtable on worldbuilding through emotional storytelling.
Same week. Same tech. Same general region. Just a wildly different vibe.
As one attendee put it on X:
“The two conferences sounded exactly the same when spoken out loud, so it was always confusing. So I just started asking: where are you — San Jose for AI, or San Francisco for Gaming?” — @joshelman

The GPU Glow Up: How Gamers Pushed GPU Innovation
Before AI was the buzzword, before Large Language Models (LLMs) started finishing your sentences, GPUs were already in the trenches: pushed to the limit by gamers who refused to settle for lag. These were the original performance purists, working for smoother rendering, higher frame rates, and photorealism years before the enterprise caught on.
They weren’t waiting for quarterly updates or a roadmap. They were water-cooling rigs, overclocking chipsets, and benchmarking at 2 a.m. It was personal. It was obsessive. And it forced GPU makers to keep up or get left behind. Load times had to drop. And anything less than ultra settings? Unacceptable. Gamers pushed hardware harder and faster than any other market. Not because they had to, but because they wanted to.
That pressure created a product that could do more than game. It laid the groundwork for today’s high-performance GPU architecture, capable of powering everything from cinematic world-building to billion-parameter AI models.
Manufacturers had to rethink performance, reengineer chips, and deliver parallel processing power that could scale under heat and pressure. The ability to process massive volumes of data simultaneously is exactly what makes GPUs so essential to AI, data science, and real-time computing today.
What started as a tool to bring digital worlds to life now powers everything from autonomous vehicles to predictive analytics. It’s no longer just about pixels. It’s about performance at scale.
And it all started with a gamer who wanted to win.
How GPUs Quietly Built AI
Before ChatGPT broke the internet, before “AI” became every startup’s middle name, GPUs were already doing the work: silently powering simulations, crunching models, and helping computer engineers build what felt, at the time, like science fiction.
In labs and back offices, GPUs were training early neural networks, modeling protein structures, and enabling self-driving car prototypes to make sense of the road. It wasn’t flashy and it wasn’t public. But it was happening.
The world met ChatGPT in late 2022, but the groundwork had been laid years before. Engineers had been building the scaffolding, training models, and wringing every ounce of performance out of GPU clusters to get here. When the general public finally showed up, the tech was ready. And it delivered.
“When GPT-3 launched, it marked a pivotal moment when the world started acknowledging this groundbreaking technology.”
— Forbes: GPT’s Major Milestones
Suddenly, the world saw what real-time inference looked like. Natural language. Instant answers. A conversation with a machine that didn’t feel like one. It was a jaw-drop moment, and it only worked because of the GPUs behind it.
What gamers once used to chase better graphics is now how the world processes thought, language, and decision-making. From massive LLMs and generative AI to real-time video generation and code synthesis, the GPU is the engine.
Here’s why it works. GPUs weren’t built for AI. They were built for graphics to process millions of pixels in parallel and render scenes in real time. But that architecture? Turns out it’s exactly what AI needs, too.
Parallelism: GPUs can run thousands of tasks at once. That’s what makes them ideal for training deep neural networks where every parameter is a math problem waiting to be solved.
Memory bandwidth: They move data fast — a necessity for working with massive datasets and models.
Low latency: Whether you’re rendering explosions or predicting the next word in a sentence, the system can’t lag.
Games tell stories in frames per second. AI tells stories in tokens per millisecond. But the demands are the same: speed, scalability, and real-time feedback.
Sourcing GPU Infrastructure
That’s the problem Inflect solves. We built a platform that surfaces real infrastructure, not estimates and not vague promises, but live availability, detailed specs, and price transparency all in one place.
You want H100s in a specific region? Filter by GPU type and location. You need colocation with dark fiber and high-density cooling? Narrow it down by network carrier, power availability, and rack specs. You want to compare pricing across three providers and book the install? Do it in a few clicks: no RFIs, no waiting for “someone to get back to you.”
Inflect isn’t a middleman. It’s a sourcing engine built for the kind of performance you can’t afford to miss. Because here’s the truth: If you’re still sending cold emails and waiting on PDF quotes, you’re already behind. The GPU gold rush is real. And it’s not just about chips, it’s about the infrastructure stack that supports them: compute, power, connectivity, space, and speed.
Inflect sits at the intersection of what GPUs used to be and what they’ve become. We connect the legacy demand that shaped them, gaming, to the future that’s scaling them: AI.
So whether you're building the next open-world hit or fine-tuning an LLM on your own cluster, this is where you find the power to do it right.
