Simulation Theory raises $2M so computers stop wasting compute resources

Simulation Theory, a start-up dedicated to reducing waste by optimizing compute resources, has successfully raised $2 million in pre-seed funding.

Simulation Theory’s technology allows businesses to leverage their existing infrastructure more efficiently, reducing cloud compute costs by up to 40% by dramatically increasing application performance.

In an interview with GamesBeat, Simulation Theory CEO Anthony Castoro said, “The company has a technology that allows people to leverage all of the power of their CPUs across any number of cores. With the rise of AI and before that Web3, the growth of requirements on resources for compute has just skyrocketed, and we see the investments that people are having to make in power that are five, ten or 15 years down the road just to meet the needs of data centers.”

He added, “Our contention is that you can’t just build your way out of that, that problem. You also need to be more efficient. There’s an opportunity in many of these cases, to reduce their compute budget by 30% to 40%.”

The funding

The round was led by Larry Russ, managing partner at Russ, August & Kabat with individual investors including Ryan Peterson, former CEO of Finger Food Advanced Technology Group and Robert Wallace of Strategic Alternatives.

The funding will be used to support further development of Simulation Theory’s innovative software development kit (SDK) designed to maximize applications’ ability to optimize existing resources to help companies save billions in overspending on hardware and cloud usage each year.

In today’s digital landscape with the widespread adoption of generative AI and complex simulations, many businesses increasingly rely on cloud services yet struggle with the skyrocketing prices associated with inefficient hardware usage.

“The Digital Revolution is over. Welcome to the Age of Optimization,” said Castoro. “As the demand for computing resources continues to skyrocket, we cannot simply build our way out of the problem. Simulation Theory is a deep technology company founded to address the fundamental computing challenges this new age presents. The Simulation Theory SDK allows customers to maximize the compute resources they already have, driving down costs, accelerating business results and promoting sustainable practices that can dramatically reduce our carbon footprint.”

Castoro said the fundraising was organic.

“We understand that creating software that scales on modern CPUs is challenging and as a result the solution has been to throw more expensive hardware at the problem,” said Randy Culley, CTO at Simulation Theory, in a statement. “Our technology makes it simple for application developers to take full advantage of multi-core CPU architectures on every popular operating system. Some of our early clients have already increased their compute performance by several orders of magnitude, reducing time to completion by as much as 90 percent on the same hardware.”

Origins

Culley and Castoro have known each other for more than 15 years. Culley was a game designer who was very focused on rendering software that could make games shine in visual effects. Around 2018, Culley started working on the multi-core problem.

“I’ve met brilliant people throughout my career in the video games industry, and Randy’s one of those people. We just keep gravitating towards each other,” Castoro said.

Culley started focusing on the problem of parallel programming for multiple CPU cores as far back as the days of the Xbox 360 and PlayStation 3, when multicore CPUs were becoming popular.

“I started working on this problem basically around the Xbox 360 and PS3 days because you couldn’t feed the GPU fast enough to keep my renderings going,” Culley said.

He branched out into full system architecture and learned how to scale cores horizontally on a local piece of hardware, whether it’s an Xbox 360 or a computer or your phone. They all have multiple cores, and very few programs use them well, he said.

While the disruption that multiple cores caused in the industry is ancient history, it had a lingering effect in that many programmers never learned how to program across multiple cores. Multithreaded software isn’t easy, and only a handful of people know how to do it in each company.

Culley sought to create software that would be able to automatically do it so that the programmers with limited knowledge didn’t have to learn code with parallelism.

Solving the problem

“Unfortunately, it’s still a really hard problem to solve, and the problem is making programs run in parallel is complicated and hard. How do you organize and synchronize events so that you don’t have stalls or crash the CPU or introduce more instability?” Culley said. “That’s really hard to do.”

And GPUs are good at parallelism, they aren’t so good at logical decision making. So CPUs are still quite necessary in everything from games to large language models for AI.

“You’re going to have to have a blend of both CPUs and GPUs,” Culley said.

The result is that the CPUs are the bottleneck and there has to be software to accelerate them by making programs that can optimize the use of the CPU cores.

“We make it so the CPUs are more fully utilized,” Culley said. “Why would you buy a 64-core CPU when the most strenuous thing out there is a video game that only uses six cores? There’s a lot of compute power that’s just sitting idle.”

Castoro said Culley solved the problem with a custom scheduler, or proprietary technology. The company has applied for a patent on the tech. It takes the form of a software development kit (SDK) that developers can incorporate into their applications, and it makes it easy to parallelize the work the app is doing without having to solve the really tricky parallel computing problems.

Visualizing the problem

It might be hard to visualize what this can do for enterprise applications. But gaming is a bit easier. If you’ve ever played a game and there is a lot happening on the screen at once — lots of soldiers in one intense battle, for instance, or a lot of explosions and movement — it puts stress on the hardware resources, which can’t keep up with the need for quick 3D rendering. Parallel code that is spread across the CPU’s cores can make this kind of scene run more smoothly.

Another example is playing a game with a split screen, where one player plays a co-op game on one side of the screen and the other player uses the other side of the screen. This is a problem that is hard because it’s like running two games at once using one game machine. The scenes on both sides of the scene show different animations and so the hardware has to render two different images at once. Multicore utilization is once again the way to solve this problem.

Such game problems are the origin of the challenge, and Culley has helped solve them. And now the company is focused on solving the same kind of parallelism problem for enterprise and cloud applications. At a time when hardware is scarce and pricey because of the demand for AI, this kind of solution from Simulation Theory is timely, Castoro said.

“With AI, there are a lot of discrete systems that you can integrate our scheduler into and parallelize within those systems. Then you can also make those systems that actually run in parallel to each other without having to have synchronization between them,” Culley said.

Customers in the enterprise

Customers including Secur3D, Encant AI, Perception Grid and Gameye are among Simulation Theory’s initial partners evaluating the benefits of a Simulation Theory technology integration in terms of future cost savings and performance gains. The company is talking to a lot of hardware vendors as well.

Secur3D, a company that moderates and safeguards UGC, is transforming how platforms, creators, and brands protect their 3D assets from infringement and unauthorized use. By leveraging Simulation Theory, Secur3D is poised to scale its operations rapidly.

“Integrating Simulation Theory will allow us to expand in ways we thought would take years,” said Nigel
Metcalf, head of product at Secur3D. “We anticipate increasing our asset intake capacity by at least 20 times and believe this technology will change how people anticipate, compute, and meet customer demand.”

Simulation Theory has also recently launched a pilot program to test the technology’s effectiveness for enterprise applications across various industries.

Simulation Theory’s mission is to solve the most complex compute problems to save companies billions. Founded in 2023 by Anthony Castoro and Randy Culley, Simulation Theory’s proprietary SDK uniquely empowers businesses to leverage existing resources efficiently and sustainably for maximum reduction in cost, increased performance and minimized impact on the environment.

In the future, the company could create software that can do a better job of spreading programs across graphics processing units (GPUs), Castoro said. In 2025, the company will release some white papers.

What’s in a name?

By the way, the company is named after the idea that we’re living in the Matrix, and we don’t know it. I asked Castoro, “Are you a believer?”

He replied, “You know, what’s the point? Either way, honestly.”

More seriously, Castoro added, “If you were to try to create a simulation that was so high fidelity that you couldn’t tell the difference between it and reality, you’d have to use all the compute resources available to you. And that’s what we enable people to do, and that’s why we selected that name.”

Perhaps the sad thing for Castoro is that the success of this startup could pull him out of gaming, at least for a while.

“I guess I am a little bit. When we were looking at conferences to go to next year, we wondered if we would go to the game conference,” Castoro said.