Business Standard ran a column on 8 August 2026 by Mihir S Sharma with a blunt thesis: artificial intelligence is "over-promising and under-arriving." Investment in AI has come to dominate financial markets and corporate spending, he argues, while the everyday experience of using it "remains, for most people, a collection of narrow conveniences and hidden algorithms."

That column is a good hook, not a good final word. It's right about the gap between AI's financial footprint and its lived impact. It understates how much has changed in exactly the places — governance, regulation, geopolitics — that matter most for a GS answer. This post checks Sharma's claims against the latest verified data, then follows the story into every paper it actually touches.

Where this shows up in the syllabus:

  • GS3 — Science & Tech / Economy: AI capex vs. daily-use gap; India's compute build-out under the IndiaAI Mission; the DeepMind math/science frontier.
  • GS2 — Governance / International Relations: India's 2026 IT Rules amendment on AI-content labelling; the India–AI Impact Summit 2026 and New Delhi Declaration; the EU AI Act as a comparator.
  • GS4 — Ethics: the trust-vs-adoption gap in clinical AI; accountability when AI-assisted decisions go wrong.
  • Essay: a fully sourced answer to UPSC's own 2024 topic, "Artificial intelligence — the god of the 21st century?"

1. The part Sharma gets right: money and usage have decoupled

Start with the financial claim, because it's the most checkable and it holds up. In the first half of 2025, AI-related capital expenditure contributed more to US GDP growth than all consumer spending combined, adding roughly 1.1 percentage points (JPMorgan Asset Management analysis, 2025). By Q1 2026, spending on AI data centres, hardware and networking had risen to about 1.4% of US GDP, up from 0.7% a year earlier. Hyperscalers — Amazon, Google, Microsoft, Meta — are on track to spend roughly $350 billion on capex in 2025 alone, a year-on-year jump in the mid-30% range.

Set against that, actual usage is modest. PwC's Global Workforce Hopes and Fears Survey 2025, drawn from nearly 50,000 workers across 48 economies, found only 14% use generative AI daily, and just 6% use agentic AI daily — the "AI that acts for you" that vendors have spent two years promoting. Sharma's number here is accurate and worth remembering: the spend is real, the daily habit mostly isn't yet.

UPSC relevance: Foundation current-affairs theme — no direct PYQ yet on AI capex/GDP linkage specifically, but it underpins the "jobless growth" and "capital-intensive vs labour-intensive growth" debate tested repeatedly in GS3 Economy (e.g. Mains 2023 GS3 on structural transformation of the economy).

2. Where it's quietly working: science, medicine, code

Sharma is also right that the frontier is real in narrow, technical domains. Google DeepMind's "AI co-mathematician," built on Gemini 3.1 and announced in May 2026, scored a new high on a benchmark specifically designed to defeat AI systems on unsolved problems; one mathematician reportedly cracked an open problem using a proof strategy the AI itself had generated. AlphaEvolve, a related DeepMind system, rediscovered and improved proofs for the finite-field Kakeya conjecture, with Gemini Deep Think verifying the logic and AlphaProof formalising the result — three AI systems closing a research loop that used to take human mathematicians years.

In medicine, the picture is more encouraging than Sharma's framing suggests. Elsevier's Clinician of the Future 2026 survey (2,757 clinicians across 118 countries) found that half of all clinicians now use AI for work, including 57% of doctors — hardly "quite low." The real story in that data is a trust and equity gap, not an adoption gap: only 41% of nurses use AI at work versus 57% of doctors, and only 42% of nurses call AI tools trustworthy today, even though 61% expect AI-assisted care to be better within five to ten years. That's a more precise, more useful fact for an answer than a blanket "AI hasn't arrived in healthcare."

UPSC relevance: Directly tested — Mains 2023 GS3, 10M: "Introduce the concept of Artificial Intelligence (AI). How does AI help clinical diagnosis? Do you perceive any threat to privacy of the individual in the use of AI in healthcare?" The Elsevier trust-gap data above is exactly the kind of verified, current evidence that upgrades a 2023-style answer if this theme resurfaces.

3. India didn't wait for the hype to settle — it hosted the summit

Here is where Sharma's "under-arrived" framing runs into a large, dated, testable fact: India spent early 2026 building AI diplomacy and infrastructure at a pace few other countries matched.

The India–AI Impact Summit 2026 was held in New Delhi (16–20 February 2026), the third in the Bletchley (UK, 2023) → Seoul/Paris (2024–25) → New Delhi lineage of global AI safety summits, under the tagline Sarvajan Hitaya, Sarvajan Sukhaya ("welfare for all, happiness for all"). Its intellectual framework rested on three Sutras — People, Planet, Progress — organised into seven thematic "Chakras": Human Capital, Inclusion, Safe & Trusted AI, Resilience, Science, Democratising AI Resources, and Social Good (PIB, February 2026). The summit closed with the New Delhi Declaration on AI Impact, a non-binding statement endorsed by 92 countries and international organisations — described by observers as the broadest multilateral consensus reached on AI to date.

Alongside the diplomacy, India scaled up its own compute base. The IndiaAI Mission, approved in March 2024 with a ₹10,371.92 crore (~$1.14 billion) five-year outlay, initially targeted 10,000 subsidised GPUs; by February 2026 it had already deployed roughly 38,000, with the government announcing 20,000 more under "AI Mission 2.0" and setting a target of 100,000 GPUs by the end of 2026 (Ministry of Electronics & IT statements, February 2026).

UPSC relevance: Prelims 2026 tested this directly — a statement-based MCQ on the AI Impact Summit's Sutras, the (non-binding) Charter for Democratic Diffusion of AI, and the New Delhi Declaration's seven Chakras. The correct Sutras are People/Planet/Progress — "Planning" is a commonly-set distractor for "Planet." Also foundational for Mains GS2 International Relations (India as a summit host/rule-shaper) and GS3 Science & Tech (compute infrastructure as industrial policy). See our Emerging Technologies notes for the full GS3 treatment.

4. Regulation is where the "under-arrived" thesis breaks down completely

If AI's real-world impact were as thin as Sharma suggests, governments would have little to regulate yet. They are regulating anyway, and on a real timetable.

India's Ministry of Electronics and IT notified the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 on 10 February 2026, effective 20 February 2026 — the most significant change to India's content-governance framework since the original 2021 IT Rules. The amendment folds "synthetically generated information" (AI-generated text, images, audio, video) into intermediaries' due-diligence duties: platforms must ensure such content is "clearly and prominently labelled" — visual labels for visual content, audio disclosures for audio — and the takedown window for unlawful content was tightened from 36 hours to 3.

The European Union's AI Act is running on a similar clock from the other direction. Its obligations phase in over four stages: prohibited practices from 2 February 2025, general-purpose AI model duties from 2 August 2025, and — as of 2 August 2026, a week before Sharma's column ran — the Act's high-risk AI system rules (Annex III) and transparency obligations (Article 50) became applicable. The final phase, covering AI embedded in regulated products, follows on 2 August 2027.

Two very different regulatory philosophies, both moving in real time: India's approach is disclosure-first (label the output, don't gate the model), the EU's is risk-tiered (gate high-risk uses, document general-purpose models). That contrast is a ready-made GS2 comparative-governance point.

UPSC relevance: No direct PYQ yet on the 2026 IT Rules amendment specifically (too recent for a Mains 2025-or-earlier answer, but exactly the kind of dated fact examiners build fresh questions around) — closest adjacent is the long-running IT Rules 2021 / intermediary liability line tested in GS2 Polity & Governance. The EU AI Act is a standard reference point in GS2 answers on global tech governance and in GS3 answers on India's regulatory approach to emerging tech. See RTI, e-Governance & Service Delivery for the governance-framework foundations this builds on.

5. The ethics layer Sharma's column skips

An adoption gap and a regulation race sitting side by side is, at bottom, an ethics problem: societies are being asked to trust systems faster than those systems earn trust. The Elsevier data above — high usage, low trust among nurses specifically — is a textbook case of the gap between capability and legitimacy that GS4 case studies are built around. So is the IT Rules labelling mandate: it exists precisely because synthetic content can now pass as real, which is an integrity-of-information problem before it is a technology problem.

The clinical-AI Mains question from 2023 already asked about privacy; a 2026-onward version of that question would reasonably extend to consent, accountability when an AI-assisted diagnosis is wrong, and who bears responsibility — the clinician, the hospital, or the model's developer. None of that requires inventing a new ethical framework. It requires applying the standard ones (utilitarian cost-benefit, deontological rights to privacy and informed consent, Rawlsian fairness in who gets access to AI-assisted care) to a live case, which is exactly what GS4 rewards.

6. The Essay paper already asked this exact question

UPSC set "Artificial intelligence — the god of the 21st century?" as a Prelims/Mains-cycle Essay topic in 2024. Sharma's column is, without saying so, a fully-formed answer to one side of that essay: AI's "godlike" claims (revolutionise work, life, cognition) are real in narrow domains and unproven at population scale. A strong essay on that topic in 2026 can now cite, with dates: the capex-usage gap (Section 1), the genuine scientific frontier (Section 2), India's summit diplomacy and compute build-out (Section 3), and the regulatory race in India and the EU (Section 4) — moving the argument from rhetorical ("is AI a god?") to evidentiary ("here is exactly how much of the promise has arrived, where, and on what timeline"). See our Technology, AI & Society essay guide for the full argument bank and quote list built around this theme.

Conclusion: the right question isn't "has AI arrived," it's "arrived where"

Sharma's framing — hype disconnected from lived reality — is true if the yardstick is a personalised AI assistant running your life. It's false if the yardstick is diplomacy, regulation, science, or the compute infrastructure states are racing to build. Both things are true at once, and the exam rewards the aspirant who can hold both: acknowledge the adoption gap PwC's own data documents, while also being able to name the New Delhi Declaration, the IT Rules amendment, and the EU AI Act's phase dates without hesitation. That combination — scepticism about the hype, precision about the facts — is what turns a current-affairs read into fifteen exam-ready marks across three papers and the Essay.

For continuing coverage as India's AI policy evolves, see our sister site Ujiyari for current-affairs digests. For the exam-oriented foundations, start with Emerging Technologies, RTI & e-Governance, and the Technology, AI & Society essay guide.

Bharat

Primary sources used in this post: Mihir S Sharma, "The delayed AI revolution," Business Standard (8 August 2026), for the claims attributed to him; PwC Global Workforce Hopes and Fears Survey 2025 (nearly 50,000 respondents, 48 economies); JPMorgan Asset Management, "Is AI already driving U.S. growth?" (2025) and related Q1 2026 AI-capex-to-GDP estimates; Google DeepMind announcements on the Gemini 3.1-based AI co-mathematician (May 2026) and AlphaEvolve/Kakeya conjecture work; Elsevier, Clinician of the Future 2026 survey (2,757 clinicians, 118 countries); PIB press releases on the India–AI Impact Summit 2026 and the "Seven Chakras" framework (February 2026) and the New Delhi Declaration on AI Impact; Ministry of Electronics and Information Technology, Gazette notification on the IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 (10 February 2026); European Union AI Act official implementation timeline (artificialintelligenceact.eu); UPSC Mains 2023 GS3 question paper; UPSC Essay paper topics list, 2024 cycle.