Solana co-founder Anatoly Yakovenko has weighed in on the intensifying artificial intelligence safety debate, arguing that calls for a slowdown in frontier AI development may be motivated by profitability concerns rather than purely safety considerations. In a brief social media post, Yakovenko wrote, “Profitability at $1 trillion mcap,” responding to Anthropic chief Dario Amodei’s recent manifesto titled “We Must Pace the Frontier.”
Amodei’s framework calls for measured development of advanced AI models, warning that AI capabilities are currently advancing faster than safety measures can address risks from autonomous systems. His proposal includes independent evaluators, coordinated standards among democratic nations, and international agreements covering model development. Amodei also cautioned that poorly controlled systems could create cybersecurity, biological, economic, and alignment risks if adequate safeguards are not implemented.
Yakovenko’s criticism highlights the financial reality facing major AI laboratories. Frontier AI development requires substantial investments in chips, data centers, skilled researchers, and electricity. As valuations climb toward trillion-dollar levels, companies face increasing pressure to demonstrate revenue rather than relying solely on investor funding and expensive infrastructure expansion. Yakovenko also joked that he had instructed Codex to conserve tokens, using humor to challenge the idea of an industrywide slowdown.
The push for AI safety has drawn support from OpenAI chief Sam Altman and xAI founder Elon Musk, creating unusual alignment among industry competitors. However, critics argue that broad regulations could strengthen dominant laboratories while increasing compliance costs for startups and open-source developers. David Sacks also challenged industrywide regulation, suggesting that concerned companies could voluntarily reduce development activities without government intervention.
Another major concern is international coordination. Governments cannot easily verify whether competing countries respect agreed development limits, and China’s participation is considered essential. Unilateral American restrictions could weaken United States competitiveness without meaningfully reducing global AI development. Investors, meanwhile, appear to expect continued infrastructure spending, even if laboratories adopt longer development and equipment procurement cycles. Slower progress could spread spending across periods while giving companies more time to convert technology into revenue.
