Anthropic CEO Dario Amodei reiterated his prediction that 'powerful AI'—his preferred term for AGI—will arrive by 2028. He describes a system smarter than top experts across most domains, with millions of autonomous agents running in parallel in datacenters.
Key facts
- AGI predicted by 2028.
- Step-function advance expected in 2026.
- Millions of agents in datacenters envisioned.
- Anthropic raised $5B in Series E for buildout.
- 'Mythos Preview' codename for 2026 model.
Anthropic CEO Dario Amodei has doubled down on his timeline for artificial general intelligence, stating in his latest article and in his book Machines of Loving Grace that 'powerful AI' will arrive by 2028 [According to @kimmonismus]. This is not a new claim—Amodei first made the prediction publicly in 2024—but the specificity of the timeline and the framing around a 'step-function advance' in 2026 are fresh details.
Amodei defines powerful AI as a system that is 'not just better than humans at individual tasks, but smarter than top experts across most important domains, able to reason, create, code, research, and act autonomously through digital interfaces.' He envisions 'a country of geniuses in a datacenter'—millions of highly capable AI agents running in parallel, working faster than humans and collaborating like an entire nation of Nobel-level experts compressed into computing infrastructure.
The Anthropic blog post references 'new models in 2028 that achieve step-function advances in capabilities (similar to the relative impact of Mythos Preview in April 2026).' The term 'Mythos Preview' appears to be a codename for a future model release, though Anthropic has not publicly confirmed this. The framing suggests that Anthropic expects a major capability jump in 2026, followed by another in 2028 that crosses the AGI threshold.
Why this timeline matters
Amodei's prediction is notable not just for its ambition but for its specificity. Most AGI forecasts from AI leaders remain vague—Sam Altman has floated 'within a decade,' while Demis Hassabis has said 'within a few years.' Amodei is one of the few to pin a precise year. The 2028 target also aligns with compute scaling trends: if training compute continues to double every 12-18 months, a 2028 model would have roughly 16x the compute of GPT-4, which some researchers argue is sufficient for AGI-level performance.
However, the prediction carries caveats. Amodei's definition of 'powerful AI' is narrower than some AGI definitions—it does not require self-awareness or general-purpose reasoning across all domains, only expert-level performance across 'most important domains.' This leaves room for debate about whether the 2028 system truly qualifies as AGI. Additionally, the 'Mythos Preview' reference implies that Anthropic expects a major breakthrough in just two years, which would require either a novel architecture, a massive increase in training data, or both.
The competitive landscape
Anthropic's timeline puts it in direct competition with OpenAI, which is reportedly targeting AGI by 2027, and Google DeepMind, which has not publicly committed to a date. The race is not just about capability but about deployment: Amodei's 'country of geniuses' vision implies massive infrastructure requirements, including datacenters capable of hosting millions of concurrent agents. Anthropic has been raising capital aggressively—including a $5B Series E in early 2026—to fund exactly this kind of buildout.
Skeptics point out that Amodei's prediction is self-serving: a 2028 AGI timeline justifies Anthropic's current spending on compute and talent, and positions the company as a leader in a market that may not materialize on schedule. But even if AGI slips to 2030 or 2032, the trajectory is clear: AI capabilities are advancing faster than most experts predicted five years ago.
What to watch
The key signal to track is whether Anthropic releases a model in 2026 that matches the 'Mythos Preview' description. If the company achieves a step-function advance in two years, the 2028 AGI timeline becomes plausible. If not, the prediction will likely be revised. Watch for Anthropic's next major model release—expected in late 2026—and whether it demonstrates the kind of autonomous reasoning and research capability Amodei describes.
What to watch
Watch for Anthropic's next major model release in late 2026, which the company has hinted will be a 'step-function advance' codenamed Mythos Preview. If it matches Amodei's description, the 2028 AGI timeline gains credibility. Also track Anthropic's datacenter spending and whether it scales to support millions of concurrent agents.
