The most anticipated software release of the last decade is out today: Grand Theft Auto 6, which Joseph Bernstein of the New York Times called “the Super Bowl of video games.” That rat b**tard got paid for a free trip to Edinburgh, where he interviewed the Rockstar team.
I’m not riffing about GTA6’s playability and story, or the enormous map (my son told me it takes 20 minutes of real time at maximum speed to traverse the map, which in video game parlance, means forever). I have only one brief point to make.
Submitted for your approval: The entire personal computer market sits on the shoulders of video games.
It always has, and it still does. Pong was the first home video game, then came Atari, Sega, and a sea of others, most of which have failed. And out of that crowd, we have the survivors: the O.G. Nintendo, of course Sony, and the 800-pound gorilla king of bloatware that makes the Xbox. Notably, the Xbox division is much less profitable than Microsoft’s other business ventures, but it makes up about 6.6 percent of global revenue.
There’s a reason Microsoft keeps it around, versus spinning it off or selling it. I’ll get to that.
There is also the throne-stealing giant, Valve, the maker of the Steam platform, which turns any PC into a video console. As Heinlein wrote, TANSTAAFL (“There’s no such thing as a free lunch”). If you have a cheap PC with cheap graphics capability and a cheap CPU, your Steam experience will be equally cheap.
I could write 2,000 words on the history of personal computers. There’s a reason the Apple ][ crushed the Radio Shack TRS-80, despite the fact that Tandy Corporation had oodles of cash back in the late 1970s and early 80s. The Apple box had color graphics. Affordable color graphics. The reason: Steve Wozniak is a hardware and software integration genius. He started at first principles, and chose each component and chip based on how much he could squeeze out of it, at the hardware level.
I’ll just say it: video game programmers are the top of the heap. They write code so close to the metal they can smell the smell the solder. Look at the history of NVIDIA. They began by making consumer video accelerator cards for base-level PCs. That market, where “enthusiasts” (addicts) build super machines with transparent cases, liquid cooling systems, and linked GPUs, still exists. Yes, it’s not a large market, but you’ll find these people camped out at Micro Center or browsing through NewEgg looking for just a few more frames-per-second, or more pixels of graphic detail.
There is a group of early adopters for video games who push the game companies to their limits, and therefore drive hardware to be ever more powerful. NVIDIA catered to this group with their GeForce models. It’s what made NVIDIA successful. But what made it mega-successful is the secret inside their GPUs. When used in a different way, a graphics processor becomes a multi-gigaflop (109 floating point mathematical operations) supercomputer on a chip. In fact, the GeForce RTX 4060 yields 15,114 GFLOPS (or 15 TFLOPS) of computational power, now simply called “compute.”
Two decades ago, Jensen Huang hired a bunch of eggheads and researchers to exploit the parallel computing and supercomputer aspects of simple graphics accelerators. And from it, we get Blackwell Ultra, which operates at 20,000,000 GFLOPS (20 petaflops) for 4-bit AI operation. That’s more compute than anyone has ever seen on a single processor before, and it’s what runs the AI revolution.
My hypothesis is simple. The entire computer infrastructure relies on gaming to advance computer science. The market is ruthless, and games are not large scale enterprise applications with sunk costs, big contracts, and executives meeting on project plans. It’s one gamer spending $79.99 for a piece of software, and immersing himself into that world for whatever number of days without showering, eating, or going to work or school. If the game isn’t worth investing time—it’s boring, difficult to play, or annoyingly laggy or graphically distracting—the core gamers won’t play it. And when the core small group of addicts won’t play, neither will the rest of us.
So bad game designers and programmers are quickly weeded out. But the best programmers are scooped up by successful companies. And you’ve got Rockstar, maker of Red Dead Redemption and Red Dead Redemption 2, along with the whole universe of Grand Theft Auto, taking 13 years to release GTA6. Thirteen years to program one masterpiece that will require the most exquisite hardware to properly run it. People are actually staying home from work or school with “GTA Flu” to play this thing.
GTA6 will drive another generation of GPUs, CPUs, monitors, projectors, and peripherals, not to mention gaming consoles, forward because gamers and designers seriously exploit and understand the hardware.
This is why Microsoft keeps Xbox around: they learn so much from the hardware and the companies that write games for it. The Xbox Series X has some pretty advanced hardware, made by AMD (NVIDIA’s competitor in the graphics market). Sony’s PS5 is even more hardcore. The rest of Microsoft, writing enterprise applications, is so far from the hardware, and has to deal with so much interoperability and “bloatware” that it’s no wonder software systems crash so often and things dependent on those systems are at risk.
If game developers wrote critical applications to run the world, I guarantee it would work better. Elon Musk believes that, and he’s a hardcore gamer—though he does tend to lie about his achievements.
A Tesla vehicle is a rolling AI machine. That kind of programming comes from the same minds that write games like GTA6. A software engineer can work for Boeing, or a defense contractor, with endless documentation and change orders, or for Tesla, or SpaceX, where iterations and pivots are almost on par with the team at Rockstar. Or they can work for NVIDIA, where the work culture is “aholic.” Many of the best engineers prefer to push the limits of hardware and software, and that makes them ideal game programmers.
I know some people lament that all that advanced hardware and programming, all those billions of dollars, the market value of giant corporations like NVIDIA, were all made rendering monsters, zombies, first-person-shooter games, and games depicting gun violence or attacking police officers. What a waste!
As the Kinks sang, “Give the People What They Want.” (H/T to my brother Jay for that one.) Gamers want an immersive experience that’s fun to play, satisfying, not headache-inducing, rewarding, and addictive. That is delivered by the GFLOP.
It turns out that 20,000,000 GFLOPS makes a pretty good platform to connect a hundred thousand nodes together as a massive neural network AI that can solve unsolved math problems, cure (or invent) diseases, eliminate (or empower) criminals, or, who knows—overwatch (or engineer) elections, be your best friend (or your enemy), and overall replace everything in your life that you have to do yourself, including driving.
As these AIs become efficient at coding themselves, with AIs writing other AIs, people are worried that we’ll get a “superintelligence.” Maybe. But we (for now) still control the machines that makes the chips, and build the computers, connect the wires and the pipes and conduits, to run these things. We should all be grateful for gaming, the one secret that propelled Elon (who plays them), Jensen Huang (who enables them) and Rockstar (which develops them) to incredible riches.
In that process, all of humanity has been enriched by the advances in computer science, and ever-increasing compute power craved by games and gamers. The day in the movie “WarGames” may be approaching faster than we believe. The WOPR figured out the futility of war by playing itself. The current crop of AIs might come to a different conclusion.
Maybe it’s best if we just all used them to play games like GTA6.



