Dubai, UAE – October 14, 2025 – In a bold declaration at one of the world's premier technology gatherings, renowned microprocessor engineer Jim Keller asserted that the future of artificial intelligence (AI) chip development is poised for a dramatic transformation. Thanks to open-source technology, creating cutting-edge AI chips will soon become significantly easier and far more affordable – provided hardware remains truly programmable and innovators are empowered to unleash their creativity.
"Actually, AI processors are simpler than people think," Keller told Anadolu Agency on the sidelines of GITEX Global in Dubai. "And people would like you to believe you need $100 billion to develop an AI processor – you don’t."
Keller's remarks, delivered amid the bustling halls of the Dubai World Trade Centre, underscore a seismic shift in the semiconductor industry. As global demand for AI surges – powering everything from autonomous vehicles to medical diagnostics – the barriers to entry are crumbling. His company, Tenstorrent, is at the forefront of this movement, pioneering open-source solutions that democratize access to high-performance computing.
The 45th edition of GITEX Global, which kicked off on Monday and runs through Friday, has drawn an unprecedented 6,800 companies from 180 countries. Billed as the world's largest tech and AI event, it serves as a crucible for innovation, where executives, startups, and governments converge to shape the digital future. Against this vibrant backdrop, Keller's insights resonated deeply, challenging the narrative of exorbitant costs and geopolitical silos that have long dominated the chip sector.
Demystifying AI Chips: Simplicity Over Hype
Keller, a 68-year-old Silicon Valley icon with a career spanning decades, has an unmatched pedigree. He spearheaded AMD's groundbreaking Zen architecture, which propelled the company from underdog to rival Intel. At Apple, he optimized chips for the iPhone and iPad, enhancing battery life and performance. His stint at Tesla focused on AI accelerators for self-driving cars. Now, as founder and CTO of Tenstorrent, he's betting big on open-source to disrupt the status quo.
"AI processors are simpler than people think," he reiterated, dismissing the industry's self-perpetuated myth of complexity. Traditional chip design, he explained, often inflates costs through proprietary tools, closed ecosystems, and layers of abstraction. In contrast, Tenstorrent's approach strips away these inefficiencies.
The company has open-sourced a comprehensive suite of technologies: from AI-specific processors to general-purpose CPUs, and even the AI compiler – the software that translates high-level code into chip instructions. This compiler, publicly available on GitHub, allows developers worldwide to tweak and optimize without licensing fees.
"Open-source chips cost less and feature a more accessible structure," Keller said. Early adopters, including research labs in Europe and startups in Asia, report development times slashed by up to 70% and costs reduced by half compared to proprietary alternatives like NVIDIA's CUDA ecosystem.
Tenstorrent's flagship Wormhole chip, for instance, delivers 2,000 teraflops of AI performance at a fraction of the price of competitors. By making hardware "really programmable," as Keller emphasized, users can reconfigure cores on-the-fly for tasks like natural language processing or image recognition. "People can be creative about how they build the new solutions," he added, envisioning a wave of bespoke AI hardware tailored to niche industries.
A Bright Horizon: No Limits in Sight for AI
Keller's optimism extends beyond hardware. "The current models are really good; they’re still getting better," he said. Despite daily headlines proclaiming AI's "limits" – from energy consumption to data scarcity – he views these as overblown. "Every day I read an article that something’s hit a limit – we haven’t got close to the limit. The demand for it is really, really big; so, I don’t know, the next five years are going to be really interesting."
This aligns with industry data. According to a recent McKinsey report, AI could add $13 trillion to global GDP by 2030, with chip demand exploding 10-fold. GITEX exhibitors showcased prototypes addressing these needs: energy-efficient edge AI for IoT devices, quantum-resistant encryption chips, and neuromorphic processors mimicking the human brain.
Tenstorrent's open-source push amplifies this momentum. By 2024, the RISC-V architecture – an open standard Keller champions – powered 15% of new AI chips, up from 5% in 2022. "It's not just cheaper; it's faster to innovate," noted Dr. Sarah Lin, a Tenstorrent partner at Stanford University, in a separate GITEX panel.
Geopolitical Restrictions: A Recipe for Catch-Up, Not Containment
Keller didn't shy away from thorny issues. Export controls on advanced chips, imposed by the US and allies on countries like China, are "not going to work out in the long run," he predicted. "Because some of the technology is really open already, (and) people publish a lot."
He cited China's rapid progress: Huawei's Ascend 910B chip now rivals NVIDIA's A100 in benchmarks, developed despite sanctions. "What turns out is (that), when you restrict one area, you just force that area to develop it by themselves, and then so far what’s happened is that (it) just means they catch up."
At GITEX, this dynamic was palpable. Chinese firms like SMIC unveiled domestic 5nm AI chips, while UAE-based G42 announced a $1.5 billion partnership with OpenAI for open-source AI infrastructure. "I hope we get to a world where it’s more open and there’s less restrictions," Keller said, advocating for collaboration over confrontation.
Experts agree. A 2025 Semiconductor Industry Association study found that open-source initiatives have accelerated non-US chip production by 40%, narrowing the gap. "Restrictions breed resilience," said GITEX keynote speaker Jensen Huang of NVIDIA, echoing Keller's sentiment.
Lessons from Titans: Culture as the Ultimate Chip
Reflecting on his journeys at AMD, Apple, and Tesla, Keller highlighted cultural variances. "Every company is kind of different, right, (and) they have different cultures."
At AMD, a scrappy underdog ethos fueled the Zen revolution, turning a $2 stock into a $150 behemoth. Apple's secrecy bred precision, yielding the M1 chip's 3.5x efficiency gains. Tesla's urgency drove Dojo supercomputers for AI training.
Yet, Keller distilled a universal formula: "The thing that works best is when the team is driving towards a really good goal, like a great product, and then I really like when the engineers can own their own work and be really inspired about what they do every day and who they work with."
At Tenstorrent, this manifests in an "open environment where everybody gets to learn, gets to do something, (and) gets to own their destiny." Flat hierarchies and hackathons have yielded 20 patents in two years, with 80% employee-driven.
GITEX attendees, from Indian coders to Saudi investors, buzzed about emulating this. A side event on "Open AI Cultures" drew 500 participants, featuring Keller's fireside chat.
GITEX Global: Epicenter of Tomorrow's Tech
The event itself is a microcosm of Keller's vision. Spanning 1.2 million square meters, GITEX features AI zones with live demos: robots composing symphonies, drones inspecting oil rigs, and holograms debating ethics. Over 180,000 visitors from 180 countries network across 30 conference tracks.
Highlights include the UAE's $100 million AI Chip Fund, targeting open-source startups, and Microsoft's pledge for 1 million RISC-V developers by 2027. "GITEX isn't just an event; it's the spark," said organizer Trixie LohMirmand.
As the week unfolds, Keller's words linger: AI chips aren't rocket science – they're open-source opportunity. In five years, he predicts, 50% of AI hardware will be programmable and public, slashing costs by 80% and igniting a creativity boom.
For now, Dubai pulses with possibility. From Tenstorrent's booth, where coders fork the compiler live, to global pavilions forging alliances, the message is clear: The chip revolution is here, and it's open to all.
