Sam Altman doesn’t hedge. The OpenAI chief executive sees artificial intelligence surpassing human intelligence by 2030. Some of his biggest rivals think it could happen even sooner. The stakes could not run higher.
Altman laid out his view in a recent interview. “I would certainly say by the end of this decade, so, by 2030, if we don’t have models that are extraordinarily capable and do things that we ourselves cannot do, I’d be very surprised,” he told the Politico. That threshold marks superintelligence for him. Systems that drive scientific discovery at speeds no human team could match. Tools that invent new algorithms, uncover novel materials, perhaps even chart paths to space colonization.
But talk is one thing. Building the machines demands another. Enter Stargate. The joint venture announced in January 2025 with SoftBank, Oracle and others aims to pour $500 billion into U.S. AI infrastructure over four years. SoftBank shoulders financial responsibility. OpenAI takes operational control. Masayoshi Son chairs the effort. Key technology partners include Microsoft, NVIDIA, Arm and Oracle. (OpenAI announcement).
The project started as a codename. It stuck. Bloomberg captured Altman walking through the details with Emily Chang in June 2025, discussing the data center push alongside humanoid robots and OpenAI’s product plans. The scale dwarfs anything that came before. Early updates show five new U.S. sites added by September 2025, pushing toward 7 gigawatts of capacity and over $400 billion committed ahead of schedule. (OpenAI update).
The Compute Hunger That Defines the Timeline
Altman has long argued that more compute unlocks the next leaps. In the Fortune article that framed his 2030 prediction, he tied superintelligence directly to the need for roughly 10 times today’s computing power. Stargate exists to deliver exactly that. Flagship facilities in Abilene, Texas, already train models like GPT-5.5 on massive NVIDIA clusters. Expansions continue despite occasional pullbacks on specific sites, as power grids strain and interconnection queues lengthen.
Yet the hardware push tells only part of the story. Altman’s personal blog post from June 2025 maps a tighter sequence of breakthroughs. “2025 has seen the arrival of agents that can do real cognitive work; writing computer code will never be the same,” he wrote. Then comes 2026 with systems that generate novel insights. By 2027, robots operating effectively in the physical world. (Sam Altman blog).
He calls the overall arc a gentle singularity. Not a sudden explosion but a smooth curve that feels impressive yet manageable in real time. “The singularity happens bit by bit, and the merge happens slowly,” Altman explained. From today’s vantage the slope looks vertical. Looking backward it flattens into steady progress. The phrase captures his optimism. Progress arrives gradually enough for society to adapt. Or so he hopes.
Others in the field push the dates forward. Elon Musk has repeatedly forecast AGI arriving in 2025 or 2026. Dario Amodei of Anthropic also sees transformative systems sooner than Altman’s outer bound. These differing views reflect genuine uncertainty about scaling limits. No one knows exactly when the next qualitative jump occurs. But all agree the direction points upward. Fast.
OpenAI itself has shifted language. Executives now speak of moving past AGI toward superintelligence. In early 2025 Altman stated the company basically knows how to build AGI as previously defined. Attention turns to what lies beyond. Internal road maps reportedly target an AI research assistant at intern level by late 2026 and a fully autonomous “legitimate AI researcher” by 2028. Chief scientist Jakub Pachocki described superintelligence as systems smarter than humans across many critical domains. (TechCrunch).
The infrastructure bet carries real risks. Some reports question whether Stargate can hit every target. Power shortages, local opposition to massive data centers, and ballooning costs have forced adjustments. One planned expansion near Abilene was scrapped after talks with Oracle collapsed. Critics point to these hiccups as signs the AI investment frenzy may face economic gravity. Yet the core facilities keep growing. Demand for intelligence keeps rising.
And the implications stretch far past code or chatbots. Altman envisions AI accelerating scientific discovery to the point that breakthroughs once measured in decades arrive in years. New computing substrates. Better training methods. Perhaps even high-bandwidth brain interfaces or practical fusion. By 2035 the cumulative effect could reshape entire industries and societies. That possibility explains the urgency behind Stargate and similar projects from competitors.
Policy conversations have begun to match the technical pace. OpenAI published a paper in June 2026 outlining ideas for an “Intelligence Age” industrial policy. The document argues incremental tweaks won’t suffice as superintelligence approaches. Instead it calls for measures to expand opportunity, share prosperity and strengthen institutions so the gains reach broad populations. (OpenAI policy paper).
Altman has also floated the need for a new social contract on the scale of the New Deal. In an Axios interview he described superintelligence as so disruptive that current economic arrangements may prove inadequate. The man steering one of the largest bets in history wants the conversation to start now. Before the systems arrive.
Skeptics abound. Some former colleagues question Altman’s consistency on safety and governance. Others doubt whether today’s neural networks can truly reach the heights he describes without fundamental architectural changes. Yann LeCun and others argue scaling alone falls short. World models, better planning, causal reasoning. These elements may prove necessary. The debate continues even as clusters expand and models improve.
Still the empirical record favors the scalers so far. Each jump in compute and data has delivered surprising gains. GPT-5 already outperforms humans on many professional tasks. Agents handle multi-step workflows. Multimodal systems process video and audio with growing fluency. The trend line holds.
So what does a world with superintelligence actually look like? Altman paints a future of abundance if aligned properly. Cheap intelligence available to everyone. Scientific progress that solves climate, disease and energy constraints. Yet he acknowledges risks. Misalignment, concentration of power, unequal distribution of benefits. Those concerns explain why OpenAI continues to discuss safety measures that may prove unpopular at release time.
The coming years will test every assumption. Will 2026 really bring systems capable of original scientific insight? Can Stargate deliver the power needed without massive delays? And if models do cross the threshold Altman sets for 2030, will society prove ready?
One thing seems clear. The race has left the station. Altman, his partners and his competitors pour resources into ever-larger training runs. They race to build the infrastructure first. To shape the rules second. And to define what intelligence means in the process.
The gentle singularity may feel smooth in hindsight. Living through the next few years promises to test that claim. Short sprints of capability gain. Longer debates over governance. Bursts of economic transformation. All against the steady hum of data centers consuming gigawatts to train the next model.
Altman bets the curve bends toward extraordinary capability within this decade. His blog, his interviews, the Stargate project itself. They all point the same direction. The question now shifts from whether to when. And what comes after.
WebProNews is an iEntry Publication