Bending Spoons buys Airtable, Atlassian AI opt-out clock ticks

Atlassian's data-for-AI clock runs out in one week. Bending Spoons buys Airtable for $1.285B — a $9B discount from peak. OpenAI cuts API prices up to 80% at the low end. The week in SaaS, through the pricing lens.

Atlassian starts training AI on your Jira data August 17 — unless you opt out

On August 17, Atlassian begins collecting customer data from Jira, Confluence, and its other cloud products to train its AI models, including Rovo — enabled by default for roughly 300,000 cloud customer organizations. The policy, first reported in April, covers two categories: metadata (story points, sprint dates, SLA values, content classifications) and in-app content (Confluence page titles and text, Jira work-item descriptions and comments).

The pricing angle is in who gets to say no. Every plan can turn off in-app content collection. But opting out of metadata collection is reserved for Enterprise customers — Free, Standard, and Premium tiers cannot. Collected data can be retained for up to seven years. Full data privacy has quietly become an enterprise-tier feature: part of what you buy at the top price point is the right not to contribute training data. Competitors noticed — GitLab is running switch campaigns on exactly this point.

This is a new pricing lever worth naming: vendors monetizing customer data as an implicit part of the subscription, with privacy sold back as an upgrade. Expect others to test the same structure.

For buyers: if you run Jira or Confluence, go to the Atlassian admin console → Settings → AI features → Data usage and toggle off data contribution before August 17.

Bending Spoons buys Airtable for $1.285B — the "buy-and-reprice" playbook lands on a household name

On August 4, Bending Spoons announced a definitive agreement to acquire Airtable for $1.285 billion enterprise value (~$2.25 billion equity value once Airtable's cash is added). It's Bending Spoons' first deal after its July 2026 Nasdaq IPO — which raised $1.68 billion specifically to fund exactly this kind of acquisition — and closes later in 2026 pending regulatory review.

The valuation gap is the story. Airtable was valued at over $11 billion in the 2021 tech boom after raising $1.4 billion in venture funding. Buying it for $1.285B enterprise represents roughly a $9B markdown from peak. That's the SaaSpocalypse arc we've been tracking since Issue #7 landing on a household-name SaaS, sold to the year's most active repricing operator.

Bending Spoons has a known playbook: acquire mature SaaS at discount, cut cost aggressively, raise prices on retained customers. Meetup went from $16.99 to $29.99 per month within 18 months of the 2020 acquisition. Evernote and WeTransfer followed similar arcs. Airtable's ARR is growing 20%+ year-over-year to about $480 million as of June 2026 — a healthy business, but one now owned by an operator whose competitive advantage is repricing.

For buyers: if Airtable is in your stack (as it is for many SaaS teams — including this newsletter's own backend), lock multi-year renewal terms at current pricing before the deal closes. Bending Spoons' repricing typically arrives 6-12 months post-close. The current $54/user/mo Business tier is likely the last "old ownership" rate you'll see.

OpenAI cuts GPT-5.6 API prices 80% on Luna, 20% on Terra — the AI price war has a new front

On July 30, OpenAI announced an 80% cut on GPT-5.6 Luna and a 20% cut on Terra, while leaving flagship Sol untouched. Luna input tokens went from $1.00/M → $0.20/M; output from $6.00/M → $1.20/M. Terra input from $2.50/M → $2.00/M; output from $15.00/M → $12.00/M. Same day, OpenAI shipped a Fast Mode option (2x speed at premium pricing) and hard spend-limit controls for API organizations.

The cuts land inside a genuine competitive war. Cheaper Chinese open-source models have been undercutting OpenAI and Anthropic at the lower end; Microsoft has been shipping cheap in-house models to pressure Anthropic; and Anthropic itself has been publicly reframing its pricing around "cost per task" and efficiency rather than raw token cost. OpenAI's move reads as defense of the mid-and-low tier where it faces the most substitutability.

For buyers: three practical implications. First, if you use Luna for background workloads (RAG summarization, extractive Q&A, cheap classification), your monthly bill just dropped 80% overnight — reduce your reserved budget. Second, Fast Mode is a real product now — worth testing on interactive workloads where latency matters more than cost. Third, the cost floor keeps falling: the SaaS Price Hub AI Calculator now shows GPT-5.6 Luna at roughly 40% cheaper than Claude Haiku 4.5 for a typical chatbot workload. Rerun your cost comparisons — the ranking probably moved this week.

Also this week

• Autodesk completed its $3.6B all-cash acquisition of MaintainX on August 3 — the largest deal in Autodesk's history. Extends the design giant into day-to-day operations budgets for factories and facilities. Fits the broader industrial-SaaS consolidation theme.
• Adobe deferred its 2026 price increase on prosumer Creative Cloud tiers — SaaStr calls it the first crack in B2B pricing power since 2022, driven by Canva and Figma undercutting at ~$15/month. Enterprise ETLA renewals still carry 8-12% increases.
• AI cloud provider Nscale agreed to acquire Anyscale for $1.65 billion on July 30, moving to own more of the AI compute stack from GPUs to orchestration.
• Schneider Electric acquired grid-intelligence vendor AiDASH for roughly $350 million on July 30 — same industrial-SaaS consolidation pattern.