AI Tools That Shut Down in July 2026: The Zero-Casualty Report
July 2026 closed with zero new AI tool shutdowns, acqui-hires, or public wind-downs in the AI Tool Graveyard database — the first flat month in our monthly reporting, after seventeen casualties in the preceding twelve months. This closed-book report explains the three-signal verification behind the zero, the four forces that produced a quiet month, and the seven categories scored on consolidation risk for the second half of 2026. It includes three infographics — a rolling twelve-month cadence chart, a category risk board, and a six-step vendor audit scorecard — plus an eight-point buyer's playbook, guidance for founders and investors, and eight detailed FAQs.
Quick answer: how many AI tools shut down in July 2026
Zero. July 2026 closed with no new AI tool shutdowns, acqui-hires, or public wind-downs in the AI Tool Graveyard database. The month is now closed and verified. It is the first flat month in our monthly reporting.
That is unusual. Over the twelve months before it, we logged seventeen notable casualties, from the Play.ht acquisition by Meta in December 2025 back through a steady drip of voice, agent, and hardware exits. Every one of those months had at least one entry. July had none.
This report explains what the zero means, how we verified it, which categories are most likely to break the calm in the second half of 2026, and exactly what buyers, founders, and investors should do with a quiet month. If you landed here asking "which AI tools died in July 2026", the honest answer is none — and the useful answer is below.
What we checked before publishing a zero
A flat month is easy to fake by simply not looking. So we publish a zero only after three independent checks come back empty.
Check one: product health. Every tool on our watch list gets a rolling automated probe. We test DNS resolution, TLS certificate validity, and the homepage HTTP status. We also match the returned HTML against a stored fingerprint so a parked domain or a "we're sunsetting" splash page cannot pass as healthy. In July, every previously active tool returned a 200 with recognizable product content.
Check two: first-party notices. We watch company blogs, RSS feeds, status pages, changelogs, and official X accounts for sunset language. No vendor we track published a shutdown, migration, or end-of-life notice in July 2026.
Check three: acqui-hire reporting. We cross-read TechCrunch, Reuters technology, and Bloomberg Technology for talent absorptions. Nothing AI-specific crossed our reporting bar in July.
A casualty gets logged only when two of the three signals agree. Three empty signals is a real zero, not a gap in coverage. You can see the live output of the same pipeline on the AI Tool Graveyard leaderboard.
Why July 2026 was quiet
Four forces stacked up. None of them is "the AI shakeout is over."
The weak field is already gone. The 2024 and 2025 consolidation wave took out most of the fragile players. Rockset went to OpenAI. Multi went to OpenAI. Adept went to Amazon. Inflection went to Microsoft. Humane went to HP and bricked its hardware. What survives is a smaller, better-capitalized field with fewer monthly failure candidates.
Announcements cluster around earnings. Big-tech quarterly results land in late July and August. Acquirers routinely hold talent deals until they can bundle them into an earnings narrative. A silent early-summer month often means a loud late-summer one.
Summer slows everything. Founders, comms teams, and enterprise buyers all take July off. Nobody wants to announce a wind-down into an empty news cycle, and nobody wants to run a migration while half the team is away.
Runway extension replaced shutdown. Since 2025, the standard response to a bad quarter has been a down round, a bridge note, or a headcount cut rather than a shutdown. Struggling tools now fade slowly instead of dying loudly. We track that separately as "abandoned" status in the tools directory.
The categories most likely to break the calm
Quiet does not mean safe. We score every category we track on consolidation risk: platform encroachment, differentiation, funding pressure, and switching costs.
Standalone voice AI (86). The most exposed category on the board. Play.ht went to Meta in December 2025, and the remaining independents compete against voice features bundled free into frontier models. Read what happened to Play.ht for how fast this category can turn, and is Play.ht dead for the zombie-product pattern that follows an acquisition.
AI wearables (81). After the Humane AI Pin collapse and the muted Rabbit R1 reception, the ambient-hardware thesis has yet to produce a durable winner. Hardware burn is unforgiving and the phone in your pocket keeps getting better at the same job.
Meeting notetakers (74). Fireflies, Otter, and Grain all sell a feature that Zoom, Microsoft Teams, and Google Meet now include by default. Our Fireflies.ai breakdown covers the squeeze in detail.
Autonomous coding agents (66). The category is real and growing, but the independents are boxed in between GitHub Copilot, Claude Code, and OpenAI Codex. See our Devin AI alternatives ranking for who is actually shipping.
AI presentation tools (58). Gamma and its peers face slide generation appearing natively inside Google Workspace and Microsoft 365. Differentiation is now design taste, which is hard to defend.
Legal AI point tools (52). Casetext was absorbed by Thomson Reuters, and the pattern is repeating around Harvey. Incumbent legal-research vendors have distribution that startups cannot buy.
Frontier chat assistants (14). ChatGPT, Claude, and Gemini are the acquirers, not the targets. Risk here is competitive, not existential.
The twelve months before the silence
Seventeen casualties landed in the year before July 2026. They fall into four clear patterns. Knowing the pattern tells you what kind of warning you will get.
Pattern one: the clean acqui-hire. The team joins a larger lab. The product gets a sunset date. Users get thirty to ninety days. Multi and Rockset both ended this way. This is the softest landing available, and it still costs you a migration.
Pattern two: the asset sale. Someone buys the code, the brand, or the patents. The service may survive under new owners, or it may not. Humane went this route and its hardware stopped working. Read the buyer's statement carefully, not the seller's.
Pattern three: the slow fade. No announcement. The changelog stops. Support replies get slower. Pricing pages freeze. The product runs, but nobody is home. We classify these as abandoned rather than shut down, and they are the hardest to spot from the outside. Our Play.ht status check walks through what a fading product actually looks like.
Pattern four: the hard stop. Servers off, refunds maybe, data gone. Rare now, because acquirers usually pay something for a team. Still the pattern that destroys the most customer value, because it comes with the least notice.
July 2026 produced none of the four. That is the whole of the news.
What actually changed between 2024 and 2026
The market did not get safer. It got quieter, for structural reasons.
In 2024, hundreds of thin wrappers around a single model API were competing for the same customers. Most had no data moat, no distribution, and no path to profit. That population burned down fast across 2024 and 2025.
By 2026, three things had changed. Frontier labs shipped the features that wrappers were selling. Enterprise buyers consolidated onto fewer vendors. And funding shifted from breadth to depth, so fewer new fragile companies entered the pool each quarter.
The result is a smaller field of stronger companies. Fewer candidates means fewer monthly casualties. It does not mean zero risk. It means the remaining risk is concentrated in the categories scored above, and it surfaces in bursts rather than a steady drip.
Five myths about a quiet month
Myth one: the shakeout is finished. It is not. Consolidation moved to a quarterly rhythm tied to earnings. Look at the trailing four quarters, not the trailing four weeks.
Myth two: no news means your vendor is healthy. Fading products publish nothing at all. Silence from a vendor is neutral at best. Check the changelog, not the press.
Myth three: acquisition protects you. It protects the team. Your workflow still breaks when the product sunsets. Every tool in our archive that was acquired eventually changed, repriced, or closed.
Myth four: enterprise plans are safe. Enterprise contracts buy notice and support. They do not buy immortality. They do buy leverage, so use it to demand export rights.
Myth five: you can migrate later. Migration takes longer than the notice period almost every time. Voice pipelines, prompt libraries, and fine-tunes do not port cleanly. Do the dry run while nothing is wrong.
What a quiet month is worth to you
The value of July 2026 is not the news. It is the breathing room.
Most vendor-risk work never happens because there is always a fire. A month with no shutdown announcement is the one window where you can inventory your stack, test an export, and write down a fallback without doing it under pressure. Teams that ran this audit before December 2025 moved off Play.ht in days. Teams that did not spent a fortnight rebuilding voice pipelines by hand.
Work the six rows above in order. The two that people skip are the two that matter: actually running an export rather than confirming the button exists, and writing the fallback vendor down somewhere your team will find it at 9am on a bad Monday.
Buyer's playbook for the rest of 2026
- Re-inventory every AI dependency. List the vendor, the owner on your side, the renewal date, and what breaks if it disappears tomorrow.
- Check each vendor's status. The leaderboard flags degraded and abandoned tools before the press does.
- Demand an export clause. No AI contract should be signed without a written data-export right that survives termination.
- Test the export. A clause you have never exercised is a promise, not a backup.
- Name a second choice per category. Voice, chat, meetings, slides, legal, code. One documented alternative each.
- Avoid multi-year lock-in with single-product startups. Annual with a clean exit beats a discount you cannot escape.
- Watch acquirer earnings calls. Absorption plans surface there weeks before the blog post.
- Re-run the audit quarterly. One quiet month is noise. Two in a row is a signal worth investigating.
What founders should read into it
A flat month is a reprieve, not a verdict.
If you are in one of the high-risk categories above, July bought you time to do three things: extend runway, sharpen the one thing your product does that a bundled feature cannot, and open acquirer conversations while you still have leverage. Every founder in our postmortem archive had a window like this and most spent it shipping features instead of fixing distribution.
The uncomfortable pattern across our database is that the tools absorbed on good terms started those conversations a year early. The ones that shut down cold started them a month late. Read why Rockset failed and why Peppertype.ai failed side by side — same market pressure, very different endings.
What investors should read into it
Pace changed. Direction did not.
Frontier labs are still absorbing three things: infrastructure (Rockset, MosaicML), talent (Adept, Inflection, Multi), and category winners (Play.ht). July's silence reflects announcement timing, not a change in appetite. Broader trackers such as CB Insights and Crunchbase News show the same slope.
The number worth watching in Q3 is not shutdown count. It is the ratio of acqui-hires to outright closures. When that ratio falls, it means acquirers have stopped paying for teams — and that is when the graveyard fills up fast.
Early warning signs to watch every week
You do not need our pipeline to spot trouble. Five checks take ten minutes a week.
- Changelog cadence. A product that shipped weekly and now ships nothing for two months is drifting. This is the earliest and most reliable signal.
- Support response time. Ask a low-stakes question. If a same-day reply becomes a four-day reply, headcount is shrinking.
- Job postings. Open engineering roles mean investment. A careers page that empties out means the opposite.
- Pricing page edits. Sudden annual-only plans, removed free tiers, or steep new minimums often precede a sale.
- Status page history. Rising incident frequency with slower resolution suggests the team maintaining the service is thinner than it was.
Log these five per vendor once a month. Two red flags on the same vendor is your cue to test the export and warm up the backup.
A worked example: how the Play.ht exit played out
December 2025 is the clearest recent case, and it explains why the audit matters.
Play.ht was a well-regarded voice tool with real customers. When the Meta deal landed, the notice period was short and the migration was not simple. Teams did not just need a new API. They needed to re-record brand voices, re-tune pronunciation dictionaries, and re-test every audio asset already in production.
Teams with a documented backup switched in days. Teams without one lost weeks. The work was identical in both cases. The difference was whether it had been scoped before the deadline existed.
That is the entire argument for spending a quiet month on preparation. See our Play.ht alternatives ranking for the shortlist those teams ended up using.
How to keep watching without waiting for us
Live status: the AI Tool Graveyard leaderboard and the full tools directory.
Context and postmortems: the blog index, plus deep dives on Character.AI, Peppertype.ai, and Kuki AI.
Buyer's guides: Claude alternatives, ChatGPT alternatives, Perplexity alternatives, Udio alternatives.
Primary sources: vendor changelogs and status pages beat every secondary report by days.
Trade press: TechCrunch AI, The Verge AI, Ars Technica.
The verdict on July 2026
July 2026 recorded zero AI tool shutdowns. The number is real, and it is narrow. It reflects a thinner field, earnings-timed announcements, and a summer lull — not the end of AI consolidation.
Treat it as a free month. Audit your stack, test one export, write down one fallback per category. Then check the August 2026 roundup when it lands, because the categories scored above have not gotten any safer while the news was quiet.
Frequently Asked Questions
How many AI tools shut down in July 2026?
Zero. July 2026 closed with no new AI tool shutdowns, acqui-hires, or public wind-downs in the AI Tool Graveyard database. It is the first flat month in our monthly reporting. Over the twelve months before it we logged seventeen notable casualties, so a month at zero stands out sharply. We publish a zero only after three independent checks come back empty: automated product health probes covering DNS, TLS, and homepage fingerprint matching; first-party monitoring of vendor blogs, status pages, changelogs, and official accounts for sunset language; and cross-reading of TechCrunch, Reuters, and Bloomberg for reported acqui-hires. All three returned nothing for July, which makes the zero a verified result rather than a gap in coverage.
Why were there no AI shutdowns in July 2026?
Four forces stacked up. First, the 2024 and 2025 consolidation wave already removed most of the fragile players, so the surviving field is smaller and better funded with fewer monthly failure candidates. Second, big-tech quarterly results land in late July and August, and acquirers routinely hold talent deals until they can bundle them into an earnings narrative, which pushes announcements into Q3. Third, the summer slowdown is real: founders, comms teams, and enterprise buyers are all away, and nobody announces a wind-down into an empty news cycle. Fourth, struggling tools now take down rounds, bridge notes, and headcount cuts instead of shutting down, so they fade slowly rather than dying loudly.
Is the AI consolidation wave over?
No. The pace changed in July, not the direction. Frontier labs are still absorbing infrastructure such as Rockset and MosaicML, talent such as Adept, Inflection, and Multi, and category winners such as Play.ht. Nothing in July suggests acquirers lost appetite. The metric worth tracking through the rest of 2026 is not the raw shutdown count but the ratio of acqui-hires to outright closures. As long as acquirers keep paying for teams, weak startups land softly. When that ratio falls, it means teams are no longer worth buying, and that is the point at which shutdown counts rise sharply rather than staying flat.
Which AI categories are most at risk in the second half of 2026?
We score standalone voice AI highest at 86 out of 100, because Play.ht already went to Meta and remaining independents compete against voice bundled free into frontier models. AI wearables score 81 after the Humane AI Pin collapse and the muted Rabbit R1 reception. Meeting notetakers score 74 because Zoom, Teams, and Meet now include transcription and summaries by default. Autonomous coding agents score 66, boxed in between Copilot, Claude Code, and Codex. AI presentation tools score 58 as slide generation appears natively in Google Workspace and Microsoft 365. Legal AI point tools score 52 against incumbents with unbuyable distribution. Frontier chat assistants score 14 because they are the acquirers, not the targets.
How does AI Tool Graveyard verify that a tool has shut down?
Three signals, two of which must agree before we log anything. Product health comes from rolling automated probes that check DNS resolution, TLS certificate validity, HTTP status, and a fingerprint match of the returned HTML against a stored copy, so a parked domain or a sunset splash page cannot pass as healthy. First-party notices come from monitoring vendor blogs, RSS feeds, status pages, changelogs, and official accounts for end-of-life language. Acqui-hire reporting comes from cross-reading TechCrunch, Reuters, and Bloomberg Technology. Live output from the same pipeline drives the public leaderboard, and individual verdicts are recorded on each tool profile in the directory.
What should buyers do during a quiet month?
Run the audit you never have time for. Inventory every AI tool in use with its vendor, internal owner, renewal date, and blast radius if it disappears. Check each vendor's status on the leaderboard. Confirm in writing that every contract carries a data-export right that survives termination. Then actually run one real export per vendor, because an untested clause is a promise rather than a backup. Name a documented second choice for each category you depend on: voice, chat, meetings, slides, legal, and code. Finally, diary the next review for thirty days out. The whole job takes roughly a day and a half across three people and it is only possible when nothing is on fire.
What should AI founders take from a zero-casualty month?
Treat it as a reprieve rather than a verdict. If you sit in a high-risk category, July bought time to do three specific things: extend runway before you need to, sharpen the one capability a bundled platform feature cannot replicate, and open acquirer conversations while you still have leverage. The consistent pattern across our postmortem archive is that teams absorbed on good terms began those conversations roughly a year before they needed to, while teams that shut down cold started about a month too late. Reading the Rockset and Peppertype.ai postmortems side by side shows the same market pressure producing very different endings.
When does the next AI Tool Graveyard trend report publish?
The August 2026 monthly roundup publishes in early September, once the month's data is complete and verified through all three signal checks. We expect it to be busier than July because quarterly results from the major acquirers land in late July and August, and several of the categories scored in this report have not become any safer while the news was quiet. Between monthly reports, the leaderboard updates continuously with live status for every tracked tool, and the tools directory carries individual verdicts. Yearly retrospectives publish each December and cover the full-year consolidation pattern, funding retreat, and largest casualties.