Custom Brain Cells Play Video Games Now. What Happens Next?
Speakers
Summary
- Biological computing (AI wetware) is emerging as a highly energy-efficient alternative to silicon, with human brain cells learning to play Doom on just 20 watts of power compared to megawatts for GPU racks.
- Researchers successfully uploaded a mapped fruit fly brain into a computer simulation, demonstrating emergent, untrained biological navigation in a digital environment.
- Despite these sci-fi breakthroughs, biological computing remains in its infancy; the cells die quickly, require strict climate control, and currently underperform standard AI models.
- The immediate reality is a severe compute and energy bottleneck for AI labs like Anthropic, which are throttling model performance due to a lack of GPUs to support 1 million new users per day.
- The market is not pricing in an AI bubble burst anytime soon, signaling continued dominance for traditional silicon GPUs and the energy infrastructure required to power them.
Trade Ideas
Clearly, the energy efficiency is the biggest threshold. We just don't have enough energy to power these chips... fully pivoting from chips to energy. That is the biggest thing in the world. A single GPU rack uses massive amounts of power, and the current grid cannot support the exponential growth of AI data centers. Because biological computing (which runs on a fraction of the power) is decades away from replacing silicon, the immediate second-order effect of the AI boom is a massive supply-demand imbalance in electricity. Utility companies and independent power producers will command massive premiums to supply baseload power to tech giants. Long the energy and utility sector as the physical bottleneck to AI expansion shifts from silicon chips to raw electricity generation. Regulatory hurdles block new power plant construction, or AI models become drastically more efficient on silicon, reducing projected power demand.
What's the roundup of this episode? NVIDIA still wins. Silicon-based intelligence is still the frontier and Jensen's going to make a lot more money. While biological computing is incredibly energy-efficient, it is currently a fragile, lab-bound science experiment. Meanwhile, top AI labs like Anthropic are actively throttling their models because they are signing up 1 million users a day and running out of compute. This immediate, desperate need for processing power guarantees that silicon GPUs will remain the undisputed bottleneck and profit center for the foreseeable future. Long NVDA as the primary beneficiary of the ongoing, unquenchable demand for silicon-based AI compute. Biological computing or quantum computing makes an unexpected, rapid leap to commercial viability, or AI scaling laws hit a wall, reducing the need for massive GPU clusters.
Amazon needs to come in and give them some more AWS access or something because they don't really have the advantage. They need more GPUs. Anthropic's Claude is experiencing explosive growth, leading to server outages and model throttling. Because Amazon is a massive backer of Anthropic and their primary cloud provider, Anthropic's desperate need to scale compute directly translates to increased AWS utilization and revenue. Long AMZN as AWS captures the massive infrastructure spend required to keep top-tier foundational models online during hyper-growth phases. Anthropic diversifies its cloud providers, or AWS struggles to procure enough GPUs to meet the demand, leading to lost revenue opportunities.
