Bhaskar Ghosh doubts AI roll-ups of India's $400B IT/BPO services industry will earn SaaS multiples or network effects because buying revenue also buys employees.
Ghosh predicts services pricing shifts from time-and-materials to outcome-based, with US sales and India delivery.
Ghosh says Databricks, Nutanix and Rubrik built core engineering in Bangalore and Hyderabad, and 8VC portfolio firms follow.
AI opens India's $400bn services industry to roll-ups
India's $400bn services and BPO industry is the next thing generative AI rewrites, because LLMs can finally handle the unstructured, multi-step work that RPA never touched. Investors are buying legacy services firms, running them and layering AI software on top, though Bhaskar Ghosh doubts these roll-ups earn software multiples or network effects since revenue arrives with employees attached. The scarce inputs are measurable outcomes, US sales motions and India's engineering depth.
Takeaways
LLMs automate the service work RPA could not. Foundation models handle unstructured data and multi-hop workflows that rule-based automation failed on. Ghosh points to managed IT tickets such as password resets and Google Drive setup, and to supply chain invoice processing. Productivity and margins rise, though whether fewer people are needed stays contested.
AI roll-ups are not PE cost-cutting. Investors buy legacy services businesses, run them, and insert AI-first software, rather than only trimming the cost structure. Ghosh is sceptical of SaaS-style multiples. Buying $100m of revenue brings 700 employees, and reaching $400m through ten more acquisitions is not a network-based act.
India's services corridor carries $400bn. Revenue flowing into India from IT, BPO and other services is about $400bn. The wider services economy, including families, schools and hospitals, is roughly $1.2tn inside a $3.7tn Indian economy. Services TAM scales with world population and dwarfs enterprise software.
Pricing shifts from hours to outcomes. Ghosh expects time-and-materials contracts to give way to outcome-based pricing. That requires services customers can migrate away from easily and measurable proof points on quality. Buyers also need tools to judge outcomes, which do not yet exist on the services side.
Sales sit in the US, delivery in India. Mid-market Indian BPO chief executives cluster in Houston and Dallas while 90 to 95 percent of employees sit in India. Ghosh expects sales, marketing and the landing motion to move increasingly to the US, with AI-enabled delivery and headcount staying in India.
India supplies core engineering for US tech. Oracle, LinkedIn, Google, Cisco, Nutanix, Rubrik, ThoughtSpot and Databricks have built core engineering teams in Bangalore and Hyderabad. Ghosh says the strongest engineers in his own portfolio, at Rocketlane, Ushur, Yugabyte and Nile, sit in Bangalore and Madras.
Where AI-native infrastructure goes next
Observability is ripe for LLM-native rebuilds. Datadog, Splunk, Elasticsearch and New Relic defined the category, and Ghosh points to Google's TimesFM for time-series forecasting and to security observability as adjacent openings.
Tap a name to go there · tap a line to see why they are linked
On this episode 2
Show 1
Data lakes will hold more semi-structured and unstructured data as enterprises squeeze more value from it. The ingestion tier and data observability become large markets, and 8VC's Aryn works there. Acryl, built on LinkedIn's open-source Data Hub, has rebranded as Data Hub on the back of AI metadata.
Deploying agents and copilots inside enterprises needs orchestration, feedback loops, compliance and governance. Ghosh names LangChain, LangFuse, LangGraph and MCP as the tooling layer forming around it.