Verdict: Three back-to-back signals show the AI race splitting into two battles — (1) sovereign ownership of models and compute, and (2) trust and control of frontier models. India just landed a real domestic champion in Sarvam; Salesforce is paying a premium to stay relevant in the agentic economy; and the US government's sudden ban of Anthropic's Fable 5 / Mythos 5 proves that model capability is now a national-security issue, not a product review.
Last verified: 2026-06-17 · Three signals to watch: Sarvam's $1.5B valuation · Salesforce's $3.6B Fin deal · Fable 5 / Mythos 5 suspended globally · Volatile facts: Funding rounds, M&A close dates, model availability, and government directives can change quickly.
What happened this week
1. Sarvam AI becomes India's newest AI unicorn at a $1.5B valuation
On June 15, 2026, Sarvam AI announced the first close of a $300 million Series B — $234 million raised so far — at a $1.5 billion post-money valuation. HCLTech is the lead strategic investor with a reported $150 million commitment, joined by Bessemer Venture Partners and existing backers Khosla Ventures and Peak XV Partners. The company had previously raised about $41 million across seed and Series A rounds.
The more important story is who wrote the check. For the first time, a major Indian IT services company is leading a large Indian foundation-model round, rather than a US tech giant or foreign VC. Sarvam co-founder Pratyush Kumar called it a sign that India is building the capital structure to create globally competitive AI companies.
Sarvam is positioning itself as a full-stack sovereign AI platform, not just a model lab. It has trained models from scratch in India, including Sarvam 105B, Sarvam 30B, and Sarvam Vision. It says its conversational platform now handles more than 2 million interactions per day, its inference platform processes 10 million API calls per day, and usage has tripled in recent months. The company also says its speech models transcribe 500,000+ hours of audio per month and its document-AI platform has digitized 35 million+ pages.
Primary source: Sarvam AI — Series B announcement
2. Adani + Jabil plan to manufacture AI data-center hardware in India
The same day, Adani Group and US-based electronics manufacturer Jabil announced a strategic alliance to build a vertically integrated AI and data-center infrastructure manufacturing platform in India. The plan is to deploy multi-gigawatt manufacturing capacity for high-density AI racks, servers, storage, networking, power-distribution units, coolant-distribution units, transformers, switchgears, and thermal-management systems.
Adani had already committed $100 billion to developing green-energy-powered, AI-ready data centers by 2035. The companies cite a global market opportunity exceeding $3 trillion over the next seven years for AI infrastructure. India's data-center capacity is projected to reach 5–8 GW by 2030.
The move is a bet that India's role in the AI economy may be bigger as an infrastructure builder than as a model builder — an angle Sarvam's funding also touches, but from the compute-and-model side.
Primary source: Adani Enterprises and Jabil — Strategic Alliance announcement
3. Salesforce agrees to buy Fin for $3.6 billion to catch the agentic wave
Salesforce signed a definitive agreement to acquire Fin (formerly Intercom), an AI customer-agent company, for approximately $3.6 billion (full breakdown). The deal is expected to close in Q4 of Salesforce's fiscal year 2027, pending approvals.
Fin's AI Agent, powered by the proprietary Apex model, resolves customer-service queries across chat, email, WhatsApp, SMS, phone, and Slack. Salesforce says the acquisition will help it serve both ends of the market: Agentforce for large enterprises that want deep customization, and Fin for smaller organizations that want a packaged, quick-to-deploy agent. Fin brings 30,000+ customers and a long-tenured AI engineering team.
The deal lands as Salesforce's agent platform, Agentforce, reported $1.2 billion in annual recurring revenue in Q1 FY27, up 205% year-over-year. That is one of the first billion-dollar agentic-AI revenue lines from a top-tier enterprise software vendor.
Primary source: Salesforce — Fin acquisition announcement
4. US government suspends Anthropic Fable 5 and Mythos 5 worldwide
On June 12, 2026, Anthropic said it received a US government export-control directive ordering it to suspend all access to Fable 5 and Mythos 5 by any foreign national, including foreign-national Anthropic employees. The company disabled the models for all customers globally to ensure compliance.
The directive arrived with little detail. Anthropic's understanding is that the government became aware of a narrow, non-universal jailbreak of Fable 5. Anthropic reviewed the demonstration and said it identified only a "small number of previously known, minor vulnerabilities" that other publicly available models can also find without a jailbreak. The company strongly disagreed with the decision and said it is working to restore access.
The incident highlights a new reality: frontier-model access can be revoked overnight on national-security grounds, even for users inside the US. It reinforces the "access vs. ownership" argument that Indian leaders like Sarvam's Pratyush Kumar have been making.
Primary source: Anthropic — Statement on Fable 5 / Mythos 5 access
Why sovereign AI suddenly matters more
Sarvam's fundraise and Anthropic's ban are two sides of the same coin. If the most capable models can be pulled offline by a single government's directive, every other country faces a sovereignty risk: its enterprises, government agencies, and startups depend on intelligence they do not control.
Sovereign AI is not about matching GPT-5 parameter-for-parameter on day one. It is about having a credible, continuously improving alternative that is governed domestically, trained domestically, and deployable without foreign permission. That is the "no-regrets option" argument — and the logic behind DeepSeek's $7.4B founder-controlled round.
What this means for you
- If you run a small business or startup: Do not build critical workflows on a single frontier model API without an exit ramp. Start testing domestic or open-weights alternatives for the 80% of tasks that do not need the absolute frontier.
- If you are in enterprise IT: Expect vendor consolidation around agent platforms. Salesforce is betting the future of CRM is autonomous agents, not dashboards. Customer service is the first beachhead; sales and ops agents are next.
- If you are a builder or founder: The infrastructure layer is where India has a real opening. Models get the headlines, but racks, cooling, power, and sovereign compute are the scarce, long-term assets.
- If you care about AI safety: The Fable 5 ban shows that governments are now treating frontier models as dual-use technology. Expect more export controls, more red-teaming requirements, and more pressure on model providers to prove their guardrails under adversarial scrutiny.
Related reading
FAQ
Q: Is Sarvam AI really building models from scratch in India? A: Yes. Sarvam says it has trained models including Sarvam 105B, Sarvam 30B, and Sarvam Vision from scratch in India. Source
Q: How much is HCLTech investing in Sarvam? A: HCLTech committed $150 million as the lead strategic investor in the Series B first close. Source
Q: What does the Adani–Jabil partnership actually make? A: The alliance plans to manufacture liquid-cooled AI racks, servers, storage, networking gear, PDUs, CDUs, transformers, switchgears, bus bars, and thermal-management systems in India. Source
Q: Why is Salesforce buying Fin? A: Salesforce wants a packaged, quick-to-deploy customer-service agent to complement its customizable enterprise Agentforce platform, and to expand its SMB and commercial reach. Source
Q: Why did the US government ban Fable 5 and Mythos 5? A: Anthropic says it received an export-control directive citing national security, apparently linked to a reported jailbreak of Fable 5. The company disagreed with the decision and is working to restore access. Source
Q: Can I still use other Anthropic models? A: Yes. Anthropic stated that access to all other Anthropic models is unaffected. Source
Discussion
0 comments