I’ve spent enough years watching technology cycles to know when the air starts to feel different. Right now that feeling is everywhere in artificial intelligence. The money is flowing, the valuations are stretching, and every conversation seems to end with the same quiet question: how much of this is real and how much is just momentum. When someone who helped build one of the most important chip companies on the planet starts talking about rollercoasters and circular financing, I pay attention.
The AI Revolution Is Real, Yet The Bubble Risk Is Too
Hermann Hauser has seen every major technology wave of the past four decades. He co-founded Acorn Computers in the late seventies, played a central role in the birth of Arm, and today invests in deep-tech companies through Amadeus Capital. That kind of vantage point is rare. When he sits down and says this AI moment will create more value than any previous technology revolution, the statement carries weight. He also makes it clear that the ride will not be smooth.
In my view the most useful part of his message is the refusal to choose between optimism and caution. He treats both as true at the same time. The technology itself is transformative. The capital structure around it has already gotten ahead of itself in places. That tension is where the interesting decisions live for anyone building, investing, or regulating in this space.
Why This Wave Feels Different From Previous Tech Cycles
Most technology shifts arrive with a clear product story. Personal computers had spreadsheets and word processors. Mobile had apps and always-on connectivity. Cloud had elastic infrastructure. Artificial intelligence is different because it touches almost every layer at once. It changes how software is written, how chips are designed, how data is stored, how energy is consumed, and how companies compete for talent.
Hauser points out that the scale of value creation could exceed anything we have seen. I tend to agree. The ability to automate cognitive work at industrial scale is not a feature. It is a new factor of production. At the same time the physical constraints are already visible. Training and running large models remains expensive. Cooling the chips that power them is becoming a first-order problem. Memory bandwidth and energy costs are forcing architects to rethink assumptions that held for decades.
That combination of enormous potential and hard physical limits is what makes the current moment feel both exciting and fragile. The companies with deep capital reserves can ride out periods of recalibration. The ones that raised money at peak optimism may find the next funding round far more difficult.
Circular Financing And The Valuation Stretch
One of the sharper observations Hauser makes concerns the recent wave of circular financing arrangements. When large players invest in each other, commit to large purchase contracts, or structure deals that effectively recycle capital inside the same ecosystem, the reported growth numbers can start to look stronger than the underlying demand. I have seen versions of this pattern before in earlier cycles. It rarely ends cleanly.
He is careful not to claim that the largest model companies will disappear. Firms with significant cash reserves and strategic partners can survive a reset in market expectations. The risk sits more with the constellation of smaller players and the secondary markets that have priced in uninterrupted exponential growth. When expectations reset, the first casualties are usually the ones that assumed capital would remain cheap and abundant forever.
This is a revolution that will create more value than probably any other technology revolution that we’ve ever seen. But it will be a rollercoaster.
That single sentence captures the dual nature of the moment better than most lengthy analyst reports. The revolution is real. The path will include sharp drops as well as climbs.
The Semiconductor Reality Check
Chip companies have been the clearest stock-market winners of the AI boom so far. Investors correctly identified them as the picks and shovels of the new era. Yet Hauser notes that something is already changing inside the industry. The current generation of architectures is running into energy and cooling walls. Moving data between processors and memory consumes a surprising share of total power. That inefficiency is no longer a minor engineering detail. It is becoming a strategic constraint.
Two emerging approaches receive particular attention: in-memory computing and photonic computing. Both aim to reduce the energy cost of data movement. In-memory designs try to perform calculations closer to where the data already lives. Photonic approaches replace some electrical pathways with light. Neither is a complete solution yet, but both represent the kind of architectural rethink that can shift competitive positions over a decade.
Hauser admits he never expected artificial intelligence to force such a fundamental change in computer architecture. Coming from someone who lived through the original Arm design philosophy, that statement lands with force. The companies that solve the energy and memory bottlenecks will likely define the next phase of the industry the way Arm once redefined mobile silicon.
Europe’s Structural Challenge In Scaling Global Champions
Europe produces strong research and creative startups. What it struggles with, according to Hauser, is the journey from promising company to true global competitor. The capital markets, the willingness to take large risks, and the density of experienced operators who have scaled before all remain thinner than in the United States. That gap is not new, yet the speed of the AI race makes it more consequential.
I have watched European deep-tech teams raise impressive early rounds only to hit a wall when they need the kind of capital that funds multi-year infrastructure buildouts. The talent is there. The follow-on funding and the industrial partners that can absorb early products at scale are harder to find. Closing that gap will require more than another set of government grants. It will require cultural and regulatory shifts that treat technological leadership as a strategic priority rather than a nice-to-have.
Technological Sovereignty Without Isolation
Hauser’s strongest warning concerns dependence. Europe remains heavily reliant on foreign suppliers for critical layers of the technology stack, from foundation models to semiconductor design software. In an era of export controls and rising geopolitical tension, that dependence carries real risk. Maintaining close partnerships with allies is essential. Becoming a technology colony is not.
The language is deliberately blunt. A technology colony does not set its own standards, does not control its most sensitive supply chains, and does not retain the ability to innovate independently when external conditions change. Europe still has time to avoid that outcome, but the window is not infinite. The choices made in the next few years about research funding, industrial policy, and capital market design will shape the continent’s position for decades.
I find the balance he strikes useful. He does not argue for decoupling or for fortress Europe. He argues for partnership with open eyes. Keep the collaboration with the United States. Build enough independent capability that the collaboration remains a choice rather than a necessity.
Where The Next Architectural Breakthroughs May Appear
Beyond the immediate energy and memory problems, Hauser points toward longer-term shifts that could prove equally significant. Quantum computing still sits further out on the horizon, yet the progress in error correction and qubit quality continues. More immediately practical are the hybrid approaches that combine classical accelerators with specialized AI silicon and new memory hierarchies.
The companies that treat architecture as a strategic variable rather than a fixed constraint are the ones most likely to capture the next wave of value. History suggests that the winners of one generation of computing rarely dominate the next without reinventing themselves. The same pattern is already visible in the current AI buildout. The firms that locked themselves into yesterday’s assumptions about power, bandwidth, and cooling are discovering the limits of those assumptions in real time.
Practical Implications For Investors And Builders
For investors the message is straightforward even if the execution is hard. Distinguish between the durable technology trend and the temporary capital-market enthusiasm. The former will create substantial value over a decade or more. The latter can reverse in a single funding cycle. Companies with strong balance sheets, real customer demand, and defensible technical advantages sit in a different risk category from those whose primary asset is narrative momentum.
For builders the lesson is equally clear. Energy efficiency and data-movement costs are no longer secondary concerns. They are becoming central design constraints. Teams that treat them as first-class problems will ship products that remain viable when power prices rise or cooling capacity becomes scarce. Teams that ignore them will discover the limits the hard way.
- Focus capital on companies solving physical bottlenecks rather than pure software scaling stories
- Watch circular financing arrangements for signs that reported growth is outrunning organic demand
- Prefer teams with experience navigating previous technology resets
- Treat European deep-tech startups as high-potential but higher-friction opportunities
- Monitor architectural research in in-memory and photonic computing as leading indicators of the next competitive shift
The Longer Horizon Beyond The Current Boom
Every technology cycle eventually settles into a more mature phase. The early exuberance fades, the weaker players exit, and the surviving companies focus on sustainable economics. Artificial intelligence will follow the same pattern. The difference this time is the breadth of impact. Once the dust settles, the tools that remain will reshape knowledge work, scientific discovery, and industrial processes at a scale that previous waves never reached.
Hauser’s perspective is valuable precisely because it refuses the false choice between uncritical enthusiasm and reflexive skepticism. The revolution is genuine. The risks of overvaluation and circular capital are also genuine. Navigating both at once is the actual work of the next several years.
I keep returning to the architectural point. The fact that AI is forcing a fundamental rethink of how computers move and process data may prove more consequential than any single model release. The companies and regions that solve those physical constraints will shape the next decade of technology in ways that are still hard to fully imagine. That is the part of the story that feels most durable to me.
What Europe Can Still Do Differently
The sovereignty conversation often slides into protectionism or resignation. Hauser avoids both. The practical path involves keeping open research collaboration while deliberately building domestic strength in the layers that matter most. Semiconductor design tools, advanced packaging, energy-efficient architectures, and the talent pipelines that support them all deserve focused attention.
Capital markets will need to evolve as well. Patient capital that can fund multi-year hardware and infrastructure projects remains scarce in Europe relative to the United States. Changing that reality requires both policy adjustments and a shift in investor expectations about acceptable timelines and risk profiles. None of this is easy. All of it is still possible.
In my experience the regions that treat technological capability as a core element of national resilience tend to fare better when external conditions tighten. Europe has the scientific base and the industrial heritage. What it needs is the sustained political and financial commitment to turn that base into enduring competitive strength.
Reading The Signals In Real Time
Market data already shows the strain points. Demand for advanced chips continues to climb, yet supply remains constrained by manufacturing capacity, packaging bottlenecks, and energy availability. Memory pricing has moved in ways that reflect the sudden surge in AI-related consumption. Cooling infrastructure for large training clusters has become a limiting factor in some locations. These are not abstract concerns. They are measurable constraints that shape what can be built and at what cost.
At the same time the largest model companies continue to invest heavily in next-generation systems. That dual reality—visible physical limits alongside continued aggressive investment—is exactly the environment in which architectural innovation tends to accelerate. The pressure creates both risk and opportunity.
| Constraint | Current Impact | Potential Response |
| Energy consumption | High training and inference costs | In-memory and photonic designs |
| Memory bandwidth | Data movement dominates power budget | New memory hierarchies |
| Cooling capacity | Limits cluster density | Architectural efficiency gains |
| Capital concentration | Circular financing risks | Focus on organic demand metrics |
The table above is a simplified snapshot, yet it captures the core tensions. Each constraint creates both a near-term headwind and a longer-term incentive for innovation. The organizations that treat the constraints as design problems rather than temporary inconveniences will be better positioned when the current capital environment normalizes.
A Personal Note On Timing And Perspective
I have watched enough cycles to know that the most useful voices are rarely the loudest. Hauser has the advantage of having built real companies and then watched those companies reshape entire industries. That experience produces a different kind of caution. It is not the caution of someone who doubts the technology. It is the caution of someone who has seen valuations detach from fundamentals before and knows how the story usually ends for the participants who ignore the detachment.
The AI wave is still young. The productivity gains are only beginning to appear in measurable form outside the largest technology firms. The physical infrastructure required to support the next phase of growth is still being built. In that context, treating every valuation as permanent and every growth projection as conservative is a form of risk itself.
At the same time, dismissing the entire movement as pure hype would be equally mistaken. The underlying capabilities continue to improve. The applications continue to expand. The companies that survive the inevitable recalibration will likely look more valuable in ten years than they do today, even if their intermediate path includes sharp corrections.
Closing Thoughts On The Road Ahead
The central message remains the one Hauser delivered with characteristic directness. This is a genuine technological revolution capable of creating extraordinary value. It is also a market environment in which some valuations have clearly gotten ahead of themselves and in which circular financing arrangements introduce additional fragility. Both statements can be true at once.
For Europe the additional layer is sovereignty. Partnership with the United States remains essential. Dependence that approaches colonial status is not. Building enough independent capability to keep the partnership a matter of choice rather than necessity is the strategic task of the coming decade.
I expect the next few years to deliver both impressive technical progress and uncomfortable market adjustments. The organizations and regions that prepare for both will navigate the rollercoaster more successfully than those that assume continuous upward momentum. That preparation begins with listening carefully to the people who have already lived through several previous versions of the same story.
The architecture of computing is shifting under the pressure of artificial intelligence. Energy, memory, and data movement are no longer secondary engineering details. They are becoming the central competitive variables. Whoever solves those variables most effectively will help determine which companies and which regions lead the next phase of the technology industry. That is the deeper story beneath the current excitement about models and valuations, and it is the story that will still matter long after today’s funding rounds have been forgotten.
