Why GPUs Become a Growth 'Tax'

What happens when the AI race stops being about size and starts being about sense?

In this episode of Tech Talks Daily, I sit down with Wade Myers from MythWorx, a company operating quietly while questioning some of the loudest assumptions in artificial intelligence right now. We recorded this conversation during the noise of CES week, when headlines were full of bigger models, more parameters, and ever-growing GPU demand. But instead of chasing scale, this discussion goes in the opposite direction and asks whether brute force intelligence is already running out of road.

Wade brings a perspective shaped by years as both a founder and investor, and he explains why today's large language models are starting to collide with real-world limits around power, cost, latency, and sustainability. We talk openly about the hidden tax of GPUs, how adding more compute often feels like piling complexity onto already fragile systems, and why that approach looks increasingly shaky for enterprises dealing with technical debt, energy constraints, and long deployment cycles.

What makes this conversation especially interesting is MythWorx's belief that the next phase of AI will look less like prediction engines and more like reasoning systems. Wade walks through how their architecture is modeled closer to human learning, where intelligence is learned once and applied many times, rather than dragging around the full weight of the internet to answer every question. We explore why deterministic answers, audit trails, and explainability matter far more in areas like finance, law, medicine, and defense than clever-sounding responses.

There is also a grounded enterprise angle here. We talk about why so many organizations feel uneasy about sending proprietary data into public AI clouds, how private AI deployments are becoming a board-level concern, and why most companies cannot justify building GPU-heavy data centers just to experiment. Wade draws parallels to the early internet and smartphone app eras, reminding us that the playful phase often comes before the practical one, and that disappointment is often a signal of maturation, not failure.

We finish by looking ahead. Edge AI, small-footprint models, and architectures that reward efficiency over excess are all on the horizon, and Wade shares what MythWorx is building next, from faster model training to offline AI that can run on devices without constant connectivity. It is a conversation about restraint, reasoning, and realism at a time when hype often crowds out reflection.

So if bigger models are no longer the finish line, what should business and technology leaders actually be paying attention to next, and are we ready to rethink what intelligence really means?

Useful Links

• Connect with Wade Myers
• Learn More About MythWorx

Thanks to our sponsors, Alcor, for supporting the show.