NVIDIA CEO Jensen Huang declared that ‘AGI has arrived’ with OpenAI’s Astra model. The statement comes amid a flurry of frontier model launches this month, with Anthropic’s Claude Fable 5.1, Google’s Gemini 3.8 Flash, and Meta’s Muse Spark 1.3 all arriving within the first week of September.
While frontier labs are pushing the boundaries of AI, the question arises whether enterprises are prepared to use these increasingly powerful models.
Databricks Senior Vice President and General Manager for Asia Pacific & Japan (APJ) Simon Davies thinks otherwise and echoes the views of Databricks CEO Ali Ghodsi on AGI. “You could say AGI is here. The AI, though, still needs the context to be relevant and impactful for your organisation,” Davies says in an exclusive interview with AIM.
“If you can’t give the AI the right context and understanding of your organisation, your structures, the way your business units work together, the way your business processes work together, and the way you support your customers and stakeholders, it’s very hard for that AI to deliver value back to your organisation.”
Davies says providing the right data context is one of Databricks’ key priorities as enterprises scale AI adoption.
Databricks’ Four AI Priorities
Databricks is seeing strong momentum. The company raised $5 billion at a $190 billion valuation, six months after its previous $134 billion valuation. The company’s annualised revenue run rate had crossed $7 billion, with revenue growing more than 80% year over year in the second quarter.
Davies identifies context, control, cost and choice as the four key pillars of Databricks’ AI strategy.
The first is context, which involves bringing enterprise data together so AI systems have the information required to produce relevant results.
The second is control and governance. This includes understanding data lineage and determining which agents and AI systems can access specific data.
Cost is becoming equally important as enterprises scale AI. Davies says companies need to move away from simply maximising token consumption and instead focus on the value generated from those tokens.
Databricks’ Unity Gateway helps enterprises manage those costs through policies, model routing and the ability to use different models for different tasks.
The fourth pillar is choice. With frontier and open-weight models evolving rapidly, Davies said enterprises need the ability to switch between models rather than be tied to a single provider. The company works with models from frontier providers such as OpenAI and Anthropic, while also supporting open-weight models.
Notably, Databricks is currently not building frontier models itself. DBRX was the last model it released, two years ago.
That choice of models is particularly relevant in India, where Davies expects enterprises to increasingly adopt open-weight models.
The reasons, according to Davies, include the scale of the Indian market, local engineering talent and the economics of serving large numbers of users.
He cites an emerging fintech that is exploring open-weight models because using foundational models for a personalised agentic experience was proving too expensive.
“We definitely see in India there’s an appetite,” Davies says, adding that India could be at the forefront of turning open-weight models into production applications.
Lakebase Emerges as Databricks’ Agentic Database Play
Databricks is also moving into transactional workloads with Lakebase, a Postgres database built to support the growing needs of agentic applications.
Davies describes Lakebase as an agentic database built for the next generation of software, where AI agents increasingly create and interact with applications.
Lakebase has already crossed a $100 million revenue run-rate, making it one of Databricks’ fastest-growing products as enterprises ramp up agentic AI adoption.
A key capability of it is database branching, which allows agents to create copies of databases and track changes without directly operationalising those changes against core production data.
“You can create copies of the database and then be able to understand and track all of the changes that the agentic applications are making without operationalising on the core data,” Davies explains.
Lakebase also uses zero-copy data, allowing analytical and transactional workloads to work from the same underlying data.
Davies believes this could become increasingly important as enterprises build agentic applications that need both transactional and analytical capabilities.
According to Davies, the biggest obstacle to enterprise AI adoption is the struggle to connect AI projects to measurable business outcomes.
“Organisations initially tried to unlock as many AI use cases as they could without really taking the time to think about the business impact it’s going to have,” he says.
Many pilots failed to generate expected value because they lacked clearly defined success criteria or a connection to the organisation's broader transformation strategy.
India Market
Databricks considers APJ its fastest-growing region globally, with Davies saying the business is growing at more than 100% year over year across its APJ markets.
India is a particularly important part of that expansion.
Databricks announced a $250 million investment in India, with plans to train more than half a million customers and partners over three years. The company now has more than 1,000 employees in India and opened an R&D centre in Bengaluru last year.
“We’re looking to continue to expand that and aggressively hire to support our customers in this market,” Davies says.
Davies cites several customer examples from India. Announced at Databricks Data + AI Summit 2026, Reliance Jio is among Databricks’ major enterprise customers in India, using the platform to bring together data across its previously siloed businesses.
Jio migrated around 25 petabytes of data to Databricks in six months, consolidating more than 12,000 tables and running over 1,500 jobs simultaneously, Davies said in an interview with SiliconANGLE.
Davies says at Zepto, an AI agent handles more than 120,000 user interactions per day, while query times have been cut by nearly 70%. Freshworks has used agentic capabilities to accelerate support operations and reduce platform maintenance costs by 75%.
Air India is using natural-language queries across its data to support operational use cases.
At IndusInd Bank, Davies said the company unified around 1.5 petabytes of data across 44 business units. Insights that previously took 30 hours were reduced to about 45 minutes, with the resulting data foundation being used to support AI-driven customer experiences.
What About Snowflake?
Despite increasingly competing across areas such as data formats, databases and AI infrastructure, Davies says Databricks and Snowflake remain both competitors and partners.
“We have some parts of our operations compete with Snowflake, but a lot of our solution also will complement Snowflake,” he says.
Databricks’ key differentiation, according to Davies, is its emphasis on open technologies and avoiding vendor lock-in.
The company supports multiple table formats, including Iceberg and Delta, while also building around open technologies such as Spark and Postgres.
The other major differentiator is total cost of ownership. Davies points to Databricks’ ability to separate storage from compute, allowing compute resources to scale independently of stored data.
Lakebase can also automatically scale compute up or down based on demand and suspend it entirely during idle periods, reducing compute costs.
For AI agents, this could become particularly significant as agents dynamically create databases, perform workloads and then shut those environments down when the task is complete.
Your reaction
Discussion
What's your take on this story?
No comments yet — be the first to share your view.