What Anthropic announced
On Monday, August 3, Anthropic confirmed that Claude models will be available with in-country inference in India "in the coming weeks" through Amazon Bedrock, AWS's managed platform for generative AI applications. "Requests sent to Claude through the India endpoint will be processed on servers in India," the company said in a statement, as reported by The Economic Times.
The move addresses what Anthropic describes as the single most frequent request from Indian organisations: the ability to use Claude without sending sensitive data outside the country. Irina Ghose, Managing Director of Anthropic India, framed it as a threshold moment: "When that data can stay in India, AI moves from pilots into the systems that matter most."
Anthropic is not first to the table. Both OpenAI and Google already offer local data processing in India, making in-country inference a baseline requirement rather than a differentiator. But Anthropic's Bengaluru footprint — an office opened in February 2026, a growing team, and enterprise clients including Axis Bank, Godrej Industries, TCS, Swiggy, and Zoho — suggests the company is playing catch-up seriously.
Why local inference matters
In-country inference is not about latency (though that helps). It is about compliance. For regulated sectors — banking, insurance, healthcare, and government — the difference between "AI in pilot" and "AI in production" often hinges on whether data can be demonstrably kept within national borders.
Anthropic is explicit about this. The company notes that Claude deployments through Amazon Bedrock carry "the audit trails and access controls that risk and compliance teams require before anything reaches production." When you combine that with local servers, the compliance story becomes defensible.
India has been pushing hard on data localisation across sectors — from the Reserve Bank of India's mandate that payment data stay domestic, to the forthcoming Digital Personal Data Protection Act. For AI providers, offering in-country inference is no longer optional; it is table stakes for the enterprise market.
Anthropic is also embedding itself in India's tech ecosystem beyond just selling APIs. The company works with the Data Security Council of India (DSCI) and open-source platforms like MOSIP and OpenG2P, which use Claude to test for vulnerabilities. Its partner ecosystem includes Indian integrators like TCS and homegrown digital natives such as Zoho, according to Deccan Herald.
The European mirror: what the EU should be asking
This is where the contrast with Europe gets uncomfortable. Anthropic serves the EU through Anthropic Ireland Limited — its EU subsidiary based in Dublin, as noted in the company's supported countries page. Claude is available in all 27 EU member states, both via the consumer app and the commercial API. But there is no equivalent EU-specific "in-region inference" announcement comparable to what India just received.
For European enterprises governed by the GDPR and, increasingly, the EU AI Act, the lack of guaranteed EU-only data processing is more than an inconvenience. GDPR requires that personal data transferred outside the European Economic Area meet adequacy standards or rely on mechanisms like Standard Contractual Clauses (SCCs). The EU-US Data Privacy Framework, adopted in 2023, provides a legal basis for transfers to the US — but if data is routed through an AWS region in, say, Oregon, even when the customer is in Berlin, that creates a risk surface many compliance officers prefer not to explain to their DPO.
The AI Act adds another layer. High-risk AI systems — and any model used in critical infrastructure, law enforcement, or biometric applications — face strict obligations. While the Act does not mandate data localisation per se, it does require transparency about where and how data is processed, which becomes much simpler to demonstrate when inference happens within EU borders.
AWS does operate multiple European regions (Frankfurt, Ireland, Paris, Stockholm, Milan, Zurich, and Spain), so the infrastructure exists. Whether Claude inference currently runs there by default, or can be configured to, is not something Anthropic has publicly clarified for EU customers — at least not with the clarity it just showed India.
Claude pricing in EUR
For European developers and businesses, knowing what Claude costs in their own currency matters. Here are the current API prices (per million tokens) converted from USD at the approximate rate of 1 USD = 0.91 EUR:
| Model | Input (USD) | Input (EUR) | Output (USD) | Output (EUR) |
|---|---|---|---|---|
| Haiku 4.5 | $1.00 | €0.91 | $5.00 | €4.55 |
| Sonnet 5 | $2.00* | €1.82* | $10.00* | €9.10* |
| Opus 5 | $5.00 | €4.55 | $25.00 | €22.75 |
| Fable 5 | $10.00 | €9.10 | $50.00 | €45.50 |
* Introductory pricing through August 31, 2026. Standard pricing thereafter: $3.00 / $15.00 (€2.73 / €13.65). Source: claude.com/pricing.
The consumer plans are straightforward: Pro at $17/month (€15.47) billed annually or $20 (€18.20) monthly; Max starts at $100/month (€91); Team at $20/seat/month (€18.20). Enterprise pricing runs at $20/seat plus usage at API rates.
The bigger picture: sovereign AI is now a competitive feature
The India announcement is part of a broader pattern. OpenAI runs inference in European Azure regions. Google Cloud lets customers select data processing regions for Vertex AI workloads. Microsoft offers EU Data Boundary. AWS itself has long let customers pin workloads to specific regions.
What is new is how explicitly AI model providers are now competing on sovereignty as a product feature. Six months ago, most AI companies treated data locality as a backend implementation detail. Now Anthropic is issuing press releases about it, and the language — "AI moves from pilots into the systems that matter most" — reveals the target audience: enterprise risk officers who cannot deploy AI without it.
For European customers, the practical question is whether Anthropic's India announcement signals a global data-residency roadmap that will soon include explicit EU-inference guarantees, or whether India received priority treatment because of its large market size and aggressive data-localisation policies. Either way, the question is now on the table for every EU enterprise currently evaluating Claude as a production AI workload.
Does Anthropic already process Claude API requests in EU AWS regions?
AWS operates multiple EU regions (Frankfurt, Ireland, Paris, Stockholm, Milan, Zurich, Spain), and Amazon Bedrock is available in some of them. However, Anthropic has not made a specific public commitment — comparable to its India announcement — guaranteeing that Claude inference requests from EU customers are processed exclusively within EU borders. Customers with strict data residency requirements should verify the setup with their AWS account team and review Anthropic's regional compliance page.
How does India's data localisation push compare with GDPR?
GDPR does not mandate that data stays inside the EU — it regulates how data can leave the EU through mechanisms like adequacy decisions and SCCs. India's approach is more direct: sectoral regulations (like the RBI's payment data mandate) require actual on-soil storage. Both regimes create demand for in-country inference, but India's blunt instrument may produce faster infrastructure investment from AI providers — while Europe relies on its regulatory complexity to achieve similar outcomes through market pressure.
Can Indian users now access all Claude models locally?
The in-country inference rollout is coming "in the coming weeks" through Amazon Bedrock's India endpoint. Anthropic has not specified which Claude models will be available at launch, but the announcement covers the Claude model family generally. Once live, Indian organisations will access Claude through Bedrock's India region endpoint with inference processed domestically.
