Can AI Keep Scaling? Exploring the Future of AI Compute
Artificial intelligence models continue to evolve at a remarkable pace, delivering capabilities that were once the realm of science fiction. However, each significant advancement in AI demands exponentially more computational power, energy, and supporting infrastructure. This raises a critical question: how sustainable is this trajectory, and how far can AI scaling continue before hitting insurmountable limits?
Cerebras Systems, a pioneering company in AI hardware, challenges the conventional wisdom that ever-more-powerful AI systems must rely on traditional chip architectures. Instead, Cerebras has focused on wafer-scale computing, an innovative approach that constructs processors using an entire silicon wafer rather than dividing it into smaller chips. This architecture is specifically optimized for the demanding workloads of modern AI.
At TechCrunch Disrupt 2026, Cerebras Systems CEO and co-founder Andrew Feldman will share insights on the future of AI scaling during his session titled Can AI Keep Scaling?. Feldman will delve into the escalating demands for compute, energy, and infrastructure, and discuss how Cerebras’ wafer-scale approach offers a fresh perspective on these challenges.
For those eager to understand the future trajectory of AI scaling, securing a ticket to Disrupt offers a unique opportunity to hear directly from a founder who has been at the forefront of redefining AI hardware. Attendees can take advantage of a special 50% discount on tickets and bring a co-founder or colleague to join the conversation.
Cerebras Challenges the Conventional AI Chip
Andrew Feldman, co-founder and CEO of Cerebras Systems, brings deep expertise in computing infrastructure, having previously co-founded SeaMicro, a microserver startup acquired by AMD in 2012. With a history of leadership roles in technology firms like Force10 Networks and Riverstone Networks, Feldman’s experience uniquely positions him to tackle the challenges of AI hardware.
Since its founding in 2015, Cerebras has pursued an audacious goal: commercializing wafer-scale computing. Traditional chip manufacturing involves slicing a silicon wafer into numerous smaller chips, but Cerebras instead uses the entire wafer as a single, massive processor. This design offers unparalleled memory bandwidth and compute density, tailored to the intense parallelism required by AI workloads.
As AI compute demands soar, Cerebras is rapidly scaling its operations. The company raised $5.5 billion in its May 2026 IPO, underscoring strong investor confidence. Additionally, it inked a multiyear deal with OpenAI to deploy 750 megawatts of Cerebras systems between 2026 and 2028—one of the largest infrastructure commitments in the AI sector. In August 2026, Cerebras unveiled the CS-4, its newest generation of wafer-scale AI hardware, further enhancing performance and efficiency.
While wafer-scale computing offers promising gains in AI performance, questions remain about whether this approach can sustainably meet the skyrocketing compute demands. Can Cerebras’ technology scale in tandem with AI’s growth? And how will the broader infrastructure—data centers, power, cooling—keep pace? Attending Disrupt provides a chance to hear these insights firsthand and understand the future of AI hardware development.
Scaling AI Means Scaling the Infrastructure Behind It
Developing faster processors is only part of the equation. The physical infrastructure supporting AI compute—data centers, power supplies, cooling systems, and manufacturing capacity—is equally critical to sustaining growth.
Cerebras is tackling these infrastructure challenges head-on. By August 2026, the company reported over 600 megawatts of data center capacity either live or under contract, slated for delivery by the end of 2027. To support this expansion, Cerebras plans to increase its manufacturing capabilities more than tenfold during 2026 alone. Furthermore, the company is establishing its first European data center capacity this year, with ambitious plans to reach 200 megawatts by the close of 2027.
This holistic approach highlights a crucial truth: AI scaling is not exclusively about chip design but also about the ecosystem that enables those chips to operate effectively. Founders, investors, and technology leaders must consider these infrastructure constraints as they strategize for the AI future.
Andrew Feldman’s session at Disrupt will contextualize these constraints, providing an experienced viewpoint on the trajectory of compute demand, the infrastructure investments required, and potential bottlenecks if current hardware paradigms are stretched too far.
Learn What Happens When Today’s Hardware Reaches Its Limits at Disrupt
Cerebras Systems embarked on its wafer-scale computing journey well before the current AI infrastructure boom, positioning itself as a pioneer with a unique vantage point on scaling challenges. As demand for AI compute accelerates, the company is simultaneously expanding its computing and manufacturing capabilities to meet future needs.
During TechCrunch Disrupt 2026, Feldman will explore the implications of growing compute, energy, and infrastructure demands for AI’s future. He will also discuss what might transpire if conventional hardware architectures can no longer keep pace with AI’s rapid advancement. This session is particularly valuable for those building, funding, or deploying AI technologies, offering insights from a founder who has spent over a decade innovating around AI hardware limitations.
Disrupt 2026 will feature more than 200 sessions across six industry stages, including roundtables and breakout discussions, with over 10,000 founders, investors, operators, and tech leaders in attendance. The event showcases 250+ speakers and more than 300 exhibiting startups, providing unmatched opportunities for networking, dealmaking, and collaboration.
Register now for TechCrunch Disrupt 2026 to gain essential insights into how long AI can keep scaling and what changes are necessary for its sustainable growth. Take advantage of a 50% discount on tickets and bring your co-founder, partner, or colleague to share in this critical discussion about the future of AI.
Here is the source link for this article.
