日本語
Sources: [1] [2] [3] [4]
2026-08-14 Morning edition
AI News Daily
Morning edition

AI News Daily 2026-08-14

Capital keeps flowing into AI platforms while speed and autonomy both advance: Databricks lands a $190B valuation, OpenAI ships a 14x-faster API tier, Anthropic makes autonomous coding the default, and DeepSeek raises prices as inference costs bite.

Today's Highlights

Corporate
Databricks raises $5B, reaching a $190B valuation
Published: 2026-08-13
Corporate
Anthropic makes Claude Code "auto mode" the default
Published: 2026-08-14
Corporate
DeepSeek raises API prices by up to 12x
Published: 2026-08-13
Model release
OpenAI unveils "Ultrafast" mode, up to 14x faster GPT-5.6 Sol
Published: 2026-08-13
Published: 2026-08-13

Databricks raises $5B at a $190B valuation

  • Data-and-AI platform Databricks closed a $5B round led by Coatue Management, lifting its valuation to $190B, up 42% from $134B in February.
  • Annual recurring revenue now tops $7B, with quarterly revenue growing more than 80% year over year.
  • New funds will go toward products including Lakebase, its database built for AI agents.
Why it matters:Enterprise data infrastructure and AI agents are converging as the core investor thesis, a useful market signal for companies weighing their own data strategy.
Valuation
$190B
BloombergCNBC
Published: 2026-08-14

Anthropic sets Claude Code "auto mode" as the default

  • Starting August 14, new Claude Code sessions for Pro, Max, and Team plans default to "auto mode," which skips step-by-step approval.
  • A built-in classifier still asks for approval on irreversible or destructive actions.
  • In a study of 1,053 paid testers, auto mode caught 89% of risky actions, compared with 13.6% for manual review.
Why it matters:Autonomous execution becoming the default for coding agents will push companies to rethink internal review processes and AI governance.
Risky-action detection rate
89%
Anthropic (Claude) official blog
Published: 2026-08-13

DeepSeek raises API prices by up to 12x

  • Starting August 17, DeepSeek revises its API pricing; V4-Pro cached-input pricing rises about 12x and output pricing about 4.5x.
  • A new peak-pricing tier during Japan-time morning and afternoon busy hours will charge double the normal rate.
  • The increase reflects rising infrastructure costs as demand for the service has grown.
Why it matters:A price hike from a low-cost Chinese model exposes the real cost of AI inference, forcing companies to recalculate their model-selection budgets.
Max input price increase
12x
NikkeiITmedia
Published: 2026-08-13

OpenAI unveils "Ultrafast" mode for GPT-5.6 Sol

  • OpenAI announced "Ultrafast," a new API tier running on Cerebras inference infrastructure that serves GPT-5.6 Sol at up to 14x normal speed, up to roughly 750 tokens per second.
  • Testing has begun with select enterprise customers in coding, commerce, and financial-research use cases.
Why it matters:Faster inference expands what's possible for latency-sensitive chat and agent applications, informing AI-adoption decisions for real-time workloads.
Speed multiplier
14x
OpenAI official blog: "Previewing Ultrafast mode"

Trend Overview

1
AI company growth is advancing on two fronts, capital raises like Databricks and major partnerships, cementing enterprise AI adoption as a core investment theme.
2
Faster inference (OpenAI's Ultrafast) and rising costs (DeepSeek's price hike) are moving in parallel, making speed and price the key selection criteria for enterprise buyers.
3
Autonomous execution by coding agents is becoming the default (Claude Code's auto mode), requiring companies to redesign their operational governance.

Wrap-up

Three things to remember

  1. Capital keeps flowing into AI infrastructure: Databricks' $190B valuation shows data-plus-agent platforms remain a top investor bet.
  2. Autonomy is expanding by default: Claude Code's auto mode raises new questions about oversight and governance.
  3. Speed and price are diverging: OpenAI pushes for lower latency while DeepSeek's price hike raises the cost of a once-cheap alternative.

What to watch next

How enterprises balance faster, more autonomous AI tools against new governance requirements and shifting inference costs in the weeks ahead.