India AI Digest — Tuesday, June 30, 2026

• The Finance Ministry's Expenditure Finance Committee cleared a ₹1.25 lakh crore outlay for India Semiconductor Mission 2.0 — about 64% above ISM 1.0's ₹76,000 crore — sending the proposal to the Union Cabinet a day after South Korea priced its own chip-and-AI program an order of magnitude higher.
• Anthropic released Claude Sonnet 5, a mid-tier model the company positions close to Opus 4.8, at introductory pricing of $2/$10 per million tokens through August 31 — the first move below the $3/$15 Sonnet price point that has held since March 2024.
• SBI Life signed an MoU with IIT Bombay to build an AI and cyber innovation hub for insurance — a regulated insurer declaring it intends to build technology rather than buy it, with no financial specifics disclosed yet.

Position movements: compute_infrastructure 0 (India), enterprise_adoption_depth 0 (India).

---

POLICY · SEMICONDUCTORS · COMPUTE · June 30, 2026

Finance Ministry panel clears ₹1.25 lakh crore for India Semiconductor Mission 2.0

The Expenditure Finance Committee under the Finance Ministry cleared a ₹1.25 lakh crore (about $13 billion) outlay for India Semiconductor Mission 2.0 on June 30, per reporting carried across YourStory, Republic World, and Daily Pioneer citing unnamed officials. The proposal now goes to the Union Cabinet for final approval. The cleared figure is about 64% higher than the ₹76,000 crore approved for ISM 1.0 in 2021. ISM 2.0 was announced in the Union Budget on February 1, 2026, with a stated scope spanning semiconductor equipment, materials, indigenous intellectual property, chip design, manufacturing, and resilient supply chains.

What this means. An EFC clearance is the step before money exists, not the money itself. The Cabinet can ratify, trim, or restructure the outlay, and until it does, ₹1.25 lakh crore is a cleared proposal, not a program. What the clearance does establish is the government's own price for the second phase — and both the size and the stated composition are informative. ISM 1.0's ₹76,000 crore went overwhelmingly to capital-expenditure incentives for fabs and packaging plants; per the same reports, the mission has so far approved 12 projects drawing about ₹1.64 lakh crore in investment — one silicon fab, two compound-semiconductor fabs, and nine assembly-and-packaging facilities. ISM 2.0's Budget-stated scope leans a layer further upstream: equipment, materials, homegrown IP, and chip design, alongside manufacturing. If the Cabinet-approved guidelines hold that composition, the second phase is a bet on the inputs to chipmaking, not just on more plants — the harder, slower part of a semiconductor ecosystem, and the part import substitution cannot shortcut.

India angle. For the AI stack specifically, the honest read is that none of this outlay buys leading-edge AI accelerator fabrication. India's approved fab and packaging projects produce mature-node logic, compound semiconductors, and packaged chips for automotive and industrial use — not the GPUs the IndiaAI Mission procures from abroad. The nearest-term AI relevance runs through the edges: packaging capacity that could eventually handle AI-adjacent silicon, power electronics for data centres, and design IP that Indian fabless teams can build against. The outlay is real industrial policy at India's scale of capital — and that scale is the point to hold onto. At about $13 billion, ISM 2.0 is roughly one-fortieth of the ~$518 billion fab-ecosystem program South Korea unveiled the previous day. The two numbers describe different theories: Korea is buying the frontier it already occupies; India is buying the foundations of an ecosystem it does not yet have. Neither number converts into the other's outcome.

Behind the news. The June 29 digest's South Korea item closed by naming the concrete signal to watch: "the next India Semiconductor Mission funding tranche or fab approval, and whether its scale moves toward the peer benchmark or stays an order of magnitude below it." The answer arrived within a day. The tranche moved up — 64% over ISM 1.0 — and stayed an order of magnitude below the peer benchmark. Both halves of that sentence are the finding.

What to watch. Cabinet ratification of the ₹1.25 lakh crore outlay, and the scheme guidelines that follow it — specifically how the allocation splits between fab-scale manufacturing incentives and the upstream equipment, materials, and IP programs named in the Budget scope.

What this is not. Not approved spending — no rupee moves until the Cabinet signs off. And not a government announcement: no PIB release accompanied the reports; the sourcing is unnamed Finance Ministry officials, and the figure should be read as well-corroborated reporting rather than settled fact.

See also:

• South Korea commits sovereign-scale capital to chips and AI, and the number reframes India's build

Source: Republic World, YourStory, and Daily Pioneer, June 30, 2026 (report-based; proposal pending Cabinet approval). → link

Confidence: Medium. The clearance and the ₹1.25 lakh crore figure are consistently reported across multiple outlets but rest on unnamed-official sourcing with no official release; the outlay is not final until Cabinet approval.

---

MODEL RELEASE · PRICING · AGENTS · June 30, 2026

Anthropic releases Claude Sonnet 5 at $2/$10 introductory pricing, below the Sonnet tier for the first time

Anthropic released Claude Sonnet 5 on June 30. The company reports substantial improvement over its predecessor, Sonnet 4.6, on agentic performance — reasoning, tool use, coding, and knowledge work — and positions the model's performance as close to Claude Opus 4.8 at lower prices. Introductory pricing is $2 per million input tokens and $10 per million output tokens through August 31, 2026, reverting to $3/$15 after that. Sonnet 5 ships as the default model on Free and Pro plans, is available to Max, Team, and Enterprise users, and is accessible via the API as claude-sonnet-5. Anthropic's framing: "Claude Sonnet 5 is built to be the most agentic Sonnet model yet. It can make plans, use tools like browsers and terminals, and run autonomously."

What this means. Two moves in one release. The first is tier absorption, now on a monthly cadence: Opus 4.8 shipped May 28, and 33 days later the mid tier arrives with performance the company positions as approaching it, at roughly 40% of the cost ($2/$10 against Opus 4.8's $5/$25 per million tokens). This is the same pattern the Sonnet line has run since 2024 — the mid tier absorbing the prior top tier's capability — but the interval keeps compressing. The second move is the price itself. $3 input / $15 output has been the Sonnet price point since Claude 3 Sonnet in March 2024, held through 3.5 Sonnet and every Sonnet since. The $2/$10 introductory rate is the first time Anthropic has gone below it — though the August 31 reversion date is the tell that this is a time-boxed trial subsidy, not a re-pricing. Whether the market lets the price go back up is a different question from whether Anthropic intends it to.

India angle. For Indian deployers the arithmetic is direct. Agentic pipelines, coding assistants, and CX automation that ran on Opus-class models for capability reasons can evaluate Sonnet 5 during the introductory window at roughly two-fifths of Opus-tier output cost ($10/M output against Opus 4.8's $25/M), and decide before the August 31 reversion — the window is effectively a subsidized migration-evaluation period, and the disciplined move is to run the evals inside it. For the Indic consumer-app cohort, whose unit economics price against ₹80–200 monthly ARPU, a near-Opus default at $2/$10 shifts what is affordable per session, at least until September. The default-model decision matters separately: India is among the largest free-tier user bases for frontier chat products, and Free and Pro users get Sonnet 5 as the default — near-Opus capability landing in Indian consumer hands without a pricing decision. What has not moved is residency: no India region ships with this release, and for regulated Indian workloads the deployment surfaces remain the ones that existed yesterday — direct API or Azure Foundry, with a US data zone as the nearest processing guarantee.

Behind the news. The May 30 digest covered the Opus 4.8 release — itself 41 days after 4.7. Sonnet 5 extends the cadence a tier down. And it lands one day after Claude went generally available in Microsoft Foundry, the enterprise-governance path into Azure estates — a cheaper near-frontier default and a new enterprise deployment surface arriving in the same week are two halves of the same distribution push.

What to watch. August 31, 2026 — whether the $2/$10 pricing survives past the stated reversion date or $3/$15 returns as announced. Secondarily, whether Sonnet 5 appears in Microsoft Foundry's model catalog and data zones, which is where the price change would meet the regulated-enterprise channel.

Source: Anthropic, "Introducing Claude Sonnet 5," June 30, 2026. → link

Confidence: High on pricing, availability, and dates (primary source); the capability positioning relative to Opus 4.8 is Anthropic's own comparison, not independently benchmarked.

---

BFSI · CYBERSECURITY · RESEARCH · June 30, 2026

SBI Life and IIT Bombay sign an MoU to build an AI and cyber innovation hub for insurance

SBI Life Insurance signed a memorandum of understanding with IIT Bombay on June 30 to establish "Bharat's AI & Cyber Innovation Hub for Insurance," a joint research centre for the insurance sector, per the company's press release. Prof. S.V. Kulkarni, Dean of R&D at IIT Bombay, and Vishal Bhatia, SBI Life's Chief Information & Digital Officer, signed the MoU. The stated scope covers artificial intelligence, cybersecurity, and quantum technologies, spanning cyber-defence research, AI-driven tools, talent development, executive education, strategic consulting, and incubation. No financial commitment, staffing plan, or timeline was disclosed.

From the room. "The next era of value creation in financial services will be led by organisations that don't just consume technology, but actively build it." — Abhijit Gulanikar, President – Operations & IT, SBI Life Insurance

What this means. An MoU is a statement of intent, and the absent specifics set this item's ceiling: no rupee figure, no headcount, no named first deliverable. What is worth chronicling is the direction of the intent. A large regulated insurer — SBI Life is among India's largest private life insurers — is stating publicly that it wants to build AI and security technology in-house rather than procure it, and is doing so through a research university rather than a vendor. The press release frames the ambition as becoming an "AI-native insurer" — SBI Life's own characterization, and one the hub's output will either substantiate or not. The IIT Bombay bench adds a thread worth noting: the institute's side of the announcement quotes Prof. Ganesh Ramakrishnan, founding director of BharatGen, the government-backed foundation-model consortium anchored at IIT Bombay — placing the hub adjacent to India's public foundation-model effort rather than in a purely commercial lane.

India angle. The context that makes this more than a routine corporate-academic MoU is regulatory. Regulated Indian finance is moving toward a regime where AI models carry board-level governance obligations — the RBI's draft model-risk guidance set that direction for banks and NBFCs in June, and insurers under IRDAI face the same direction of travel. In that regime, an institution that builds and understands its own models holds a structurally easier compliance position than one running opaque vendor systems it must govern but cannot inspect. In-house research capability is one rational response, and a joint IIT hub is a capital-light way to acquire it. If the model works — a regulated financial institution co-running a standing research centre with a top-tier institute — it is a template other Indian BFSI players can copy, which is precisely how SBI Life's release pitches it.

Behind the news. This lands six days after the RBI released its draft model-risk management guidance (covered in the June 27 digest), which would require board-approved governance for AI/ML models across regulated finance. The institutional response side of that arc — regulated entities building capability to meet the coming governance bar — is thin so far; this MoU is an early instance, and for the insurance sector specifically it is the first such standing research hub the archive has recorded.

What to watch. The first named deliverable out of the hub — a published tool, dataset, or production deployment inside SBI Life — or a disclosed funding figure. Until one of those appears, this remains an MoU.

Source: SBI Life Insurance press release via PRNewswire, June 30, 2026. → link

Confidence: Medium. Single-source company press release; the MoU, signatories, and scope are self-reported by SBI Life and IIT Bombay, and no financial or operational specifics have been disclosed.

---

Position movements