AI models: a new contender emerges
DeepSeek's R1 — trained for ~$5.6M — wiped nearly $1 trillion off the market in a day. The Sputnik moment, the chip fallout, and tighter export controls.
"Fool lies here who tried to hustle the East," wrote Kipling. In late January 2025 the tech world and global markets were taken aback by DeepSeek, a Chinese AI startup, and its R1 model — an assistant that rivals the leading models in performance at a fraction of the development cost.
Remarkably, DeepSeek achieved this using roughly 2,000 Nvidia H800 GPUs over 55 days, for a budget of about $5.58 million. That efficiency challenges the prevailing notion that cutting-edge AI requires vast resources.
Market repercussions
R1's debut sent shockwaves through markets. Major companies including Nvidia, Microsoft and Alphabet saw significant stock declines, collectively losing nearly $1 trillion in market value — the largest market drop on record. It prompted a market-wide reevaluation of investment strategies and headlines dominated by intensifying global competition.
Navigating new challenges in chips
The semiconductor sector is grappling with the implications. Nvidia, a dominant force in AI chips, took a substantial hit on concerns about future demand for high-end chips, while Intel pursued strategic initiatives — new AI-focused chips and significant cost-cutting — to regain ground. Apple, interestingly, saw its share price rise, as its ability to deploy larger, complex models to iPhone processors looked more within reach off the back of the R1 news.
Stricter export controls
Under the new Trump presidency, the US Department of Commerce's Bureau of Industry and Security introduced stringent export controls on advanced AI technologies — covering advanced computing integrated circuits and related software — aimed at regulating the global diffusion of AI capabilities, particularly around national security.
Multimedia AI evolves at pace
Getty Images and Shutterstock announced a merger valued at around $3.7 billion to enhance their visual content and integrate AI. ElevenLabs, hot off its $180M Series C, introduced auto-regeneration to streamline workflows. And CaptionHub shipped a busy Q4 — CaptionHub Connect, a YouTube Connector, Language Weaver as a translation option, and editor enhancements like redo and timecode display.
The period between December 2024 and February 2025 saw cost-effective models like R1 disrupt the paradigm again — prompting both excitement and apprehension. A Sputnik moment, perhaps. But has China really outsmarted the US on AI? That's the question that'll define the year.