FULL STORY

DeepSeek Ends Its Low-Price Era With Up to 1,100% API Price Hike

After shaking the industry with rock-bottom API prices, DeepSeek reportedly resumed an $8 billion funding round and warned users of steep hikes. From August 16 it raised API prices by up to 1,100% with peak/off-peak billing, ending its bargain era.

2026-08-02 ~ 2026-08-16 · 5 episodes · 42 posts

Episode 1 · DeepSeek's Low Pricing Sparks Debate, Developers Question Rivals' Markups (2026-08-02, 5 posts)

DeepSeek's extremely low API pricing has sparked heated discussion among developers, with its cache-read cost being about 10x cheaper than rivals, prompting a re-evaluation of industry pricing strategies. Developer @teortaxesTex argues that comparing token prices alone is misleading, as other providers charge more partly due to privacy premiums like zero data retention (ZDR), while DeepSeek trades low margins for market share.

Confirmed

  • User @MiraMira complained that many API providers claim to match DeepSeek's prices but are actually about 10x more expensive.
  • Tech blogger @stochastichimp noted that frontier AI labs face a severe cost crisis, highlighted by DeepSeek's low prices.

Unconfirmed

  • @teortaxesTex suggests DeepSeek's low prices stem from engineering optimizations like MoE architecture, low-precision weights, and aggressive KV-cache optimization, as well as a low-margin strategy, but this remains speculative without official data.

Why it matters

  • DeepSeek's low pricing is not a simple subsidy but a result of extreme engineering capability, forcing other model providers to rethink their pricing and cost structures.

Episode 2 · DeepSeek Resumes $8B Funding, V4-Pro Specs Leak (2026-08-06, 8 posts)

DeepSeek is reportedly advancing a new large-scale funding round, with investor materials leaked online revealing its capital needs and key specs for the next-gen model V4-Pro. Multiple sources say the round has resumed, seeking $8B (or ¥50B) at a pre-money valuation of ¥500B (or $74B per Polymarket), with signing expected in late August. The leaked memo claims V4-Pro trails Anthropic's flagship by only 0.3% in coding, supports 1M-token context, costs 1/50th the price, and is adapted for Ascend. Polymarket traders see a 72% chance of IPO by 2027, but nothing is officially confirmed. If true, this would be one of the largest and highest-valued AI funding rounds, and V4-Pro's cost-performance and domestic chip adaptation could disrupt the current top-tier model landscape.

Confirmed

  • Funding progress and size: Multiple sources report DeepSeek has resumed its second round, paused earlier due to a leaked recording of founder Liang Wenfeng's meeting. The round seeks $8B (or ¥50B) at a pre-money valuation of ¥500B (or $74B per Polymarket), with signing expected in late August.
  • Use of funds: Primarily for data center construction, expansion of 10K-GPU clusters, and R&D of next-gen omni-modal models.
  • V4-Pro specs: The leaked memo claims V4-Pro will be top-tier globally, trailing Anthropic's flagship by only 0.3% in coding, supporting 1M-token context, priced at 1/50th of Anthropic's flagship, and adapted for Ascend.

Unconfirmed

  • IPO timeline: Polymarket data shows traders assign a 72% probability of IPO by end-2027, but this is market speculation, not official.
  • Material authenticity: The leaked documents are unsigned, and data has not been officially confirmed.

Why it matters

If true, this would be one of the largest and highest-valued AI funding rounds. The leaked V4-Pro specs suggest extreme cost-performance and domestic chip adaptation, which could disrupt the current top-tier model market if realized.

Episode 3 · DeepSeek Announces Significant API Price Hike, Signaling End of Low-Cost Era (2026-08-06, 25 posts)

DeepSeek has officially notified API users via banner or email of an imminent significant price increase, advising developers to plan usage accordingly. Continued use after the billing change implies acceptance of new terms; users may opt out. The exact hike and effective date are undisclosed. The increase is attributed to server overload from surging demand and compute market supply-demand tension. Current prices are $0.14/$0.28 per million tokens (input/output), far below competitors.

Confirmed

  • DeepSeek has notified API users of a significant price increase, with continued use implying acceptance of new terms.
  • The increase is due to server overload from surging demand and compute cost pressures.
  • Current prices are $0.14/$0.28 per million tokens, far below competitors.

Unconfirmed

  • Specific increase amount and effective date have not been announced; final pricing awaits official notice.

Why it matters

  • Jiqizhixin notes DeepSeek previously promised price cuts in H2; this reversal signals the end of its ultra-low-cost subsidy strategy.
  • svenai and lxfater argue that many AI products and workflows' high profits rely on extremely low API costs; a significant hike could threaten wrapper developers and small teams lacking core moats.
  • firstadopter says the hike reflects sustained strong demand for compute, breaking the bearish narrative on Nvidia demand.
  • teortaxesTex emphasizes DeepSeek's 0731 model's cost-performance gives it room to raise prices even 4x; the company rejects subscriptions, preferring to allocate compute to training and research, hence using price to curb API demand.

5 more related posts →

Episode 4 · Developer Analysis Shows Rented GPUs Can Cover DeepSeek API Costs (2026-08-07, 2 posts)

Developers recently analyzed the economics behind DeepSeek's API pricing, revealing that even with rented GPUs, the company's extremely low prices can still be profitable. This suggests their previous pricing strategy was not operating at a loss.

Episode 5 · DeepSeek Raises API Prices Up to 1100%, Shifts to Peak-Hour Billing (2026-08-14, 2 posts)

DeepSeek will raise API prices by up to 1100% starting August 16, adopting peak-hour billing while other AI labs race to cut prices. The move narrows its historical low-cost advantage.