HSBC warns India's data-centre build-out could add $105 billion to the import bill
HSBC Global Research estimates that India's planned jump from roughly 1.5 GW to 6.5 GW of data-centre capacity will require about $105 billion of imported equipment, close to 2.5 per cent of GDP. Chips account for the bulk of it, and the bank urges a staggered expansion funded by foreign capital.
India's ambition to become a regional hub for AI compute comes with a balance-of-payments price tag. A new note from HSBC Global Research puts the import requirement for the country's planned data-centre expansion at roughly $105 billion, equivalent to about 2.5 per cent of GDP. The analysis assumes capacity rises from today's 1.5 GW to around 6.5 GW as projects now under construction and in the pipeline come online.
The arithmetic is stark. HSBC estimates that each gigawatt of capacity costs about $30 billion to build, and roughly $21 billion of that is spent on imported hardware, cooling, electrical gear and replacement parts. Semiconductors alone account for around $90 billion of the projected bill. India has about $70 billion of projects under way and a further $90 billion in planning.
Because equipment is paid for up front while revenue from cloud, inference and data services arrives later, the bank cautions that the build-out could widen the current-account deficit in the near term. It does not, however, see the pressure as unmanageable: foreign direct investment, overseas borrowing and hyperscaler capital can absorb much of it.
HSBC's practical recommendation is sequencing. Rather than commissioning several gigawatts at once, operators should stagger construction so that earlier phases are generating cash before later phases draw down imports. That also leaves room for domestic manufacturing of racks, power distribution and cooling to mature, a theme reinforced by recent joint-venture announcements in the sector.
Why it matters
The note reframes India's compute story from a purely technological race into a macroeconomic one. Policymakers who want sovereign AI capacity must also plan for the foreign exchange, grid and manufacturing capacity to support it. For enterprise buyers, it signals that domestic GPU pricing is likely to remain tied to global supply for several years.
Indicops view
For government and enterprise programmes, this argues for being deliberate about where workloads sit. Most public-sector use cases, from document processing to citizen-facing assistants, are inference-heavy and can be served on modest, right-sized clusters or empanelled national compute. We help clients separate the few workloads that genuinely justify dedicated hardware from the many that do not, which keeps capital tied up in outcomes rather than idle racks.
More in AI News
Back to the newsroomJensen Huang steps down as Nvidia CEO; CFO Colette Kress becomes interim chief
Nvidia's founder resigned as chief executive on 12 November after a board meeting, with CFO Colette Kress taking over on an interim basis while a search begins. The shares fell as much as 13 per cent, and quarterly results are due on 17 November.
Google adds TPU 8tx and 8i Max, taking its eighth-generation TPU family to four chips
Google has expanded its eighth-generation TPU line with TPU 8tx, for confidential inference and small-scale training, and TPU 8i Max, a dual-chip package for long-context inference. It also open-sourced TorchTPU and TPU Sync. Availability dates for the new chips were not given.
Iran-linked hackers sabotage refrigeration controllers across Israel, the US and Europe
NPR reports that a group calling itself Cyber Isnaad Front, which Israeli incident responders attribute to Iranian military hackers, has sabotaged industrial refrigeration systems from a food plant in Israel to roughly 300 French supermarkets and, in early November, retailers across six more European countries. Internet-exposed Danfoss controllers were a common entry point.