Daisuke Okanohara is CEO of Matlantis Coporation, and Co-founder and CTO of Preferred Networks, Inc.
Every conversation about the future of AI circles back to the same assumptions: bigger models, more data, more compute. We treat progress as a contest of intelligence. Yet the constraint that will decide how far AI goes over the next decade has little to do with intelligence and a great deal to do with physics.
AI is running into two physical walls that no amount of clever software can talk its way past.
The First Wall: Energy
In a modern AI system, far more energy goes into moving data around than into performing the calculations themselves. The calculations inside a chip are remarkably cheap. The real expense lies in shuttling enormous volumes of data between memory and processor, billions of times every second. By some engineering estimates, that movement can consume ten to a hundred times more energy than the computation it serves.
For two decades, processing power has raced ahead while the speed of feeding data to it has crawled, and today’s large language models are limited less by how fast they can think than by how fast they can be fed.
The bill lands on the balance sheet and on the power grid. Running the AI we expect to deploy across the economy could require tens of gigawatts of additional electricity. The ceiling on AI’s growth is becoming an energy ceiling, and that energy is spent in a place few people think to look: the distance data must travel.
The Second Wall: Materials
The second wall sits one level deeper, in the materials our hardware is built from.
Better chips, better batteries, better energy systems: nearly every hardware advance the AI era depends on eventually comes down to discovering or improving a material. Materials discovery has always been agonizingly slow. The number of possible combinations is effectively limitless, while the number humanity has tested remains vanishingly small.
That mattered less when hardware progress had room to run elsewhere. It matters enormously now, because materials are precisely where both walls have to be broken: new architectures to shorten the distance data travels, new materials to move electricity with less waste, new chemistries to store energy more densely.
Why Spending Alone Won’t Clear Them
The instinct is to spend through these walls, buying more chips and standing up more data centers. That buys time without changing the trajectory, because it keeps optimizing the very designs that created the bottleneck.
The more durable path is using AI to redesign the physical layers beneath AI itself.
For most of computing history, those layers were designed in isolation. Materials scientists handed properties to device engineers, who handed components to chip designers, who handed hardware to system builders. When the binding constraints are physical and cut across every layer, the layers have to be designed together. Having spent years building both AI chips and the AI tools that design materials, I see this cross-layer approach as the defining shift of the coming decade.
Three changes follow in practice. Organizationally, teams that once optimized their own slice hold a shared goal and work across team boundaries, so a system-level question like halving energy per query gets answered partly at the materials level. In process, sequential handoffs become joint development, with materials treated as design variables from the start and simulation as a standard gate before capital commits to hardware. In budget, funding follows the genuine bottleneck on the path to the goal.
Materials discovery is where this change is furthest along. Understanding how a material behaves once meant either a physical experiment or a simulation too slow to model anything realistic. AI has broken that trade-off by learning the rules of how atoms interact, reaching near-laboratory accuracy at speeds that were recently impossible. At Matlantis, we have watched work that would once have demanded roughly 20 years of conventional calculation finish in about a week. Companies, including Volkswagen and Panasonic, now use this kind of simulation to explore materials at a pace no traditional lab could match.
Speed of that magnitude changes decisions by moving materials questions earlier. Screen candidates in parallel with product design instead of ahead of it. Verify early whether materials that are scarce, regulated, or exposed to supply chain risk can be replaced, and whether promising but untested ones hold up. Make go or no-go roadmap calls on simulation evidence in weeks instead of lab evidence in years.
What This Means For Leaders
Picture where this leads. AI agents survey the research, propose candidate materials and run thousands of simulations in parallel around the clock, sending only the rare promising candidate to an automated lab for physical confirmation. The bottleneck shifts from human hours to the genuine limits of physics and experiment, a far better problem to have.
Leaders can prepare now with three moves. First, build an in-house team that can read AI results critically, because the value of ten thousand simulations depends on the judgment applied to the ten that matter. Second, restructure the development cycle around that speed, with simulation running from day one rather than after prototypes exist. Third, choose partners that span layers: the suppliers, foundries and research institutions that own the parts of the stack you don’t.
A caveat: even far more capable AI will never fully solve the physical world, because reality holds layers of complexity that cannot be predicted from the rules beneath them. The useful question is where irreversibility begins. Trust AI to explore, rank and explain at a scale humans cannot. Reserve for humans every decision that cannot be undone: synthesizing a material, committing capital, taking a product to market. Experiment keeps the final word. AI changes how fast and how intelligently we can run it.
The decisive edge over the next decade will belong to whomever understands that AI’s progress now runs through the physical world, and who designs across every layer of the stack rather than optimizing a single one. Where physics sets the limit, AI is becoming the way through.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?