Is AI pricing leading to a second DeepSeek moment? - Asian Tech Roundup

Plus: Apple dispute’s India’s authority in antitrust case

Image:
Is AI pricing leading to a second DeepSeek moment?

Welcome to Computing's weekly roundup of tech news in Asia. This time we look at frontier model economics, Apple’s push back against India’s competition watchdog and Korea’s defiance of US interference in the Coupang case.

When DeepSeek first burst onto the global AI scene in late 2024, the so‑called “DeepSeek Moment” was not just about open weights or impressive reasoning benchmarks. It was about economics. An unknown Chinese startup with frontier‑level capability apparently did not need to rely on US infrastructure.

Eighteen months on, DeepSeek’s latest pricing moves raise a serious question: are we on the cusp of a second DeepSeek moment, this time driven primarily by cost?

DeepSeek V4 Pro's benchmarks put it in the same bracket as Anthropic’s Claude Opus 4.7, OpenAI’s latest GPT 5.5 and Google’s Gemini 3.1 Pro line, with roughly comparable results across coding, reasoning, and general intelligence.

Yet its pricing is dramatically lower. DeepSeek V4 Pro is priced at roughly $1.74 per million input tokens and $3.48 per million output tokens, before discounts - already a fraction of what rivals charge for equivalent models. That gap is currently even wider with DeepSeek offering 75% off API pricing until 5th May.

For comparison, Claude Opus 4.7 is priced at around $5 per million input tokens and $25 per million output tokens, GPT 5.5 is $5/$30 and Gemini 3.1 Pro comes in at $2/$12. Even without discounts, DeepSeek massively undercuts its rivals.

The contrast highlights a widening divide in the AI market. US‑based AI startups are widely reported to be losing large sums on inference, propped up by venture capital and cloud credits rather than sustainable margins. As a result, they have limited room to cut prices, and are instead capping token usage or throttling heavy users. Cheaper tokens simply translate into greater losses for companies already under pressure from investors to show a path to profitability.

DeepSeek’s position (and to an extent that of other open models such as Kimi and GLM) is different. Its open‑model strategy, access to lower‑cost infrastructure, and willingness to trade margin for adoption allow it to use pricing to its advantage.

Just as the first DeepSeek moment forced the industry to reconsider assumptions about openness and performance, this second phase challenges assumptions about who can afford to compete at the frontier.

Australia

China

India

Japan

Malaysia

Singapore

South Korea

Taiwan

Elsewhere in Asia