Baidu CEO Warns AI Bubble to Crush 99% of Startups

Baidu CEO Declares AI ‘Bubble’ Will Kill 99% of Startups

Robin Li, Baidu’s CEO, called the AI sector an “inevitable bubble” at the Harvard Business Review Future of Business Conference. He warned that 99% of AI startups will fail when the market corrects, drawing parallels to the 1990s dot-com crash. “Probably one percent of the companies will stand out,” Li said, framing the collapse as a necessary filter. Unlike past bubbles, this one hinges on AI’s ability to produce reliable outputs—specifically, to stop generating hallucinations in large language models (LLMs). Li claimed the problem has been “solved” in the last 18 months, a stance that ignores ongoing industry critiques about model accuracy in real-world use.

The CEO’s timeline for AI to displace human labor is equally optimistic. Li estimates 10–30 years before significant job losses occur, urging governments and organizations to prepare. His comments contrast with immediate concerns about economic displacement in countries like Papua New Guinea, where a software glitch sparked riots. The Baidu leadership’s focus on long-term AI risks risks overshadowing present-day tech failures.

Payroll Glitches Trigger Riots in Papua New Guinea

A seemingly minor payroll error triggered a state of emergency in Papua New Guinea. The government’s tax system miscalculated 2024 pay packets, reducing government workers’ salaries by $100—about half their monthly income. Misinterpreted as a tax hike, the error led to strikes and subsequent riots. Prime Minister James Marape called the issue a “technical glitch,” but the fallout exposed how fragile trust in digital infrastructure remains in regions with limited tech oversight.

The incident highlights a recurring theme in tech governance: the gap between system designers and end users. Software misconfigurations in critical systems—payroll, banking, or healthcare—can have disproportionate human impacts. PNG’s crisis mirrors earlier outages at Australia’s Commonwealth Bank, where a duplicated transaction error drained customer accounts. Both cases reveal a lack of redundancy and user communication in mission-critical platforms.

Asia Tightens Content Controls, Blocks X Over Deepfakes

Malaysia and Indonesia have blocked access to X (formerly Twitter) for failing to curb non-consensual sexual deepfakes. Malaysia’s Communications and Multimedia Commission demanded safeguards against illegal content, but the platform’s response fell short. Indonesia’s minister of communications echoed these concerns, framing deepfakes as a violation of human rights. Elon Musk has dismissed the bans as censorship, but the governments’ actions align with broader regulatory trends in Asia.

India’s recent rollback of a pre-approval AI licensing rule shows a more nuanced approach. The government replaced strict government permission with operational requirements—labeling deepfakes, preventing bias, and avoiding illegal content. This shift reflects India’s attempt to balance innovation with accountability. Yet the focus on deepfakes underscores a global priority: curbing synthetic media that exploits trust in digital spaces.

AI’s Regulatory and Competitive Tangle

Baidu’s decision to spin off its Kunlunxin chip business signals a strategic shift. By listing the unit separately, Baidu aims to attract investors focused on AI hardware, a sector dominated by NVIDIA and AMD. The move follows similar strategies by Alibaba and Huawei, suggesting chipmakers will dominate AI’s next phase. Meanwhile, China’s crackdown on online slang and forced labor in the supply chain complicates its role in global AI markets.

The DJI-Department of Defense dispute adds another layer. The Chinese drone maker attributes U.S. import restrictions to a “customs-related misunderstanding,” but critics point to U.S. concerns about surveillance in Xinjiang. This tension mirrors broader tech wars, where geopolitical concerns override technical merit. The outcome could shape export controls for AI and robotics, not just drones.

What to Watch

Three developments will define AI’s near-term trajectory. First, whether Baidu’s Kunlunxin spin-off attracts enough investment to rival NVIDIA. Second, how PNG’s government resolves the payroll system overhaul to prevent future crises. Third, the resolution of DJI’s lawsuit against the U.S. Department of Defense—whose stance on Chinese tech could influence global supply chains. Each case reflects the dual-edged nature of AI: transformative potential, but fragile implementation.

Updates

• 2026-07-28 — Here’s how to avoid being trapped in never-ending Apple Upgrade leases (source)