Which Business Models AI Cannot Easily Absorb, by the Numbers
Assessment data on 10,525 companies (July 2026) shows which business models resist AI absorption: chips, physical networks, regulated data, and distribution.
The question, framed by data
People ask which businesses AI can swallow and which it cannot. WillItSurvive assessed 10,525 companies in July 2026, and two verticals in that assessment answer the question almost by contrast alone: the AI companies themselves, and the retail and consumer conglomerates that AI barely touches. Reading the two side by side shows what absorption looks like and what stops it.
Of the 263 AI companies assessed in July 2026, 170 were rated losing its protection and only 2 were rated hard to displace. Of the 206 retail and consumer conglomerates assessed in the same July 2026 window, 95 were rated hard to displace. The industry closest to the technology is the one most exposed to it. The industries that own shelves, trucks, stores, and factories mostly are not.
What gets absorbed: the thin layer above the model
Inside the AI vertical, the dividing line is where a company's value sits relative to the foundation model. Companies selling a feature on top of general models keep getting the same feature shipped natively by OpenAI, Anthropic, Google, or Microsoft. The assessment already records the endpoint of that dynamic several times over, as of July 2026:
- Phind, a developer search tool, shut down in January 2026 after ChatGPT, Claude, and Gemini absorbed its niche.
- Humanloop's team was absorbed by Anthropic.
- Rewind AI was acquired and shut down by Meta. Covariant became a shell after Amazon hired its founders in 2024. Contextual AI's team was folded into a larger lab the same way.
- Soul Machines, an avatar maker, entered receivership while commoditized rivals like Synthesia and HeyGen picked up its enterprise customers.
The same absorption runs through China. 4Paradigm's commodity LLM business fell 70% year over year in Q3 2025 as Alibaba, Baidu, Tencent, and Ant Group bundled cheaper models (Qwen, Ernie, Spark, DeepSeek) into their own enterprise platforms. Mobvoi's revenue fell 17.2% overall and about 41% in the second half as voice demand shifted to iFlytek, Huawei, and Doubao. Chatbot, avatar, prompt-tooling, and standalone vision or voice vendors sit at the exposed end of the July 2026 distribution: of the 263 AI companies assessed, 88 were rated as facing high competitive pressure.
What resists absorption: below and around the model
The few strong positions inside the AI vertical share a trait: their value lives in things a model upgrade cannot erase.
- Chips and hardware. Nvidia and Broadcom rated among the strongest positions in the July 2026 assessment: Nvidia still runs the chip and software stack behind most AI training and inference, and Broadcom makes the custom AI chips hyperscalers rely on while also selling enterprise software through VMware. The watch-item here, dated July 2026, is hyperscaler in-house chips (TPU, Trainium, Maia, MTIA) slowly moving inference workloads off third-party GPUs.
- Physical operations and regulatory position. Waymo rated as the strongest position of any AI company tracked in the July 2026 assessment, on the strength of paid robotaxi service in more than 10 US cities. Driven miles and regulatory approvals do not get replicated by a chatbot feature.
- Deep customer entanglement and exclusive data. Applied Intuition, one of only 2 companies rated hard to displace among the 263 AI firms assessed in July 2026, sells simulation software that carmakers and defense agencies depend on. Dataiku is the other. PathAI holds ground in pathology image analysis, a specialized, high-trust niche with no reported customer defections. Even here nothing is unconditional: Palantir, rated defensible for now, saw Salesforce's Missionforce win a $5.6 billion Army contract in January 2026, breaching its government stronghold. That is a dated fact readers can track through Palantir's next reports.
The strongest evidence: businesses AI barely registers against
The retail vertical shows what a genuinely hard-to-absorb model looks like. Its sector pattern has almost nothing to do with AI and everything to do with who controls distribution. Walmart runs more than 10,750 stores and lost only modest grocery share, roughly 21.1% to 20.4% between 2024 and 2026 per the July 2026 assessment. Woolworths, Wesfarmers, Aeon, Reliance Retail, Costco, and CP All all carry built-to-last standings in the July 2026 assessment. Coca-Cola's bottling network and Sanrio's 450-plus licensed characters are moats no model competes with.
The retailers under pressure are being squeezed by other distribution, not by AI: Best Buy against Amazon's roughly 27-30% US electronics share versus its own 23%; CarMax against Carvana's 52% year-over-year used-vehicle revenue growth to $6.43 billion, nearly matching CarMax's $6.39 billion; QVC against TikTok Shop and Amazon Live; Fossil against the Apple Watch, with its own brand down 17% and leather goods down 41% in Q1 2026; Solo Stove down 39% from peak after clones from Breeo, BioLite, and Tiki.
The pattern in one sentence
Business models built on physical assets, exclusive or regulated data, entrenched distribution, or manufacturing scale showed low pressure in the July 2026 assessment. Business models whose product is a layer of software convenience above someone else's general-purpose model showed the highest pressure and account for most of the shutdowns and acquihires the assessment records. We never declare an outcome for any single company; the standings above reflect the July 2026 assessment, and the dated items (Palantir's contract loss, 4Paradigm's next quarter, hyperscaler chip rollouts) are things to check, not conclusions.
Questions this data answers
Which business models does the data show AI absorbing fastest? Thin software layers on top of foundation models: developer search, prompt tooling, avatars, standalone chatbots, and copywriting tools. Of the 263 AI companies assessed in July 2026, 88 were rated as facing high competitive pressure, and Phind, Humanloop, Rewind AI, Covariant, and Soul Machines had already shut down, been acquihired, or entered receivership.
Which business models resist AI pressure best? Physical distribution, chips, regulated operations, and exclusive data. Waymo's position in the July 2026 assessment reflects paid robotaxi service in over 10 US cities, and 95 of the 206 retail conglomerates assessed in July 2026 rated hard to displace because groceries and home basics still reward scale and logistics.
Are any AI companies themselves rated hard to displace? Only 2 of 263 in the July 2026 assessment: Applied Intuition, whose simulation software is embedded with carmakers and defense agencies, and Dataiku. Both hold customer entanglement that a model upgrade does not erase.
Is retail immune to technology pressure? No, but the pressure comes from other retailers, not from AI. Best Buy trails Amazon's roughly 27-30% US electronics share, and Carvana's used-vehicle revenue grew 52% year over year to $6.43 billion against CarMax's $6.39 billion, per the July 2026 assessment.
What should readers watch to test this pattern? Three dated signals from the July 2026 assessment: whether more wrapper startups get acquihired instead of raising rounds, whether Chinese hyperscalers keep bundling AI into enterprise platforms, and whether hyperscaler in-house chips expand internal deployment. Each would confirm the layer above the model keeps collapsing while the layers below it hold.
Assessments reflect evidence available July 2026. Companies change; treat watch-items as things to verify, not conclusions.
Figures are as of each company's assessment date unless dated otherwise.