Editorial: The market’s AI fever is meeting two blunt forces this week: hard financial scrutiny and the physical fragility of cloud infrastructure. Between a pulled $30bn IPO and drone strikes that knocked a major Russian cloud offline, the current moment is less about slogans and more about logistics, financing and risk.

In Brief

Microsoft confirms its upcoming Edge feature collects user data but not for spying

Why this matters now: Microsoft Edge’s new AI feature will collect diagnostic and usage data, which affects millions of browser users deciding whether to enable AI tooling in their workflows.

Microsoft confirmed that an AI‑powered feature in the Edge browser will gather telemetry intended to “keep the service up to date, secure, and operating properly,” while insisting the data collection is diagnostic not surveillance, per reporting on the feature. Browser telemetry is routine, but the combination of local browsing behavior and AI prompts raises fresh privacy questions — especially when vendors fuse browsing context with model inputs.

“Microsoft Edge collects diagnostic data to keep the service up to date, secure, and operating properly,” the company wrote in privacy documentation, according to coverage.

If you run Edge, check the privacy toggles before enabling new AI features: many diagnostic options are optional and can be narrowed to reduce what Microsoft receives.

Source: Neowin report on Edge

McKinsey: Automation could absorb half of U.S. work hours by 2035. The bigger challenge is skills

Why this matters now: McKinsey’s estimate that automation could replace the equivalent of half of U.S. work hours by 2035 puts pressure on employers and policymakers to rapidly scale retraining and credential programs.

The McKinsey analysis highlighted by Fortune argues that automation and AI might displace or transform an enormous share of current tasks — not necessarily whole jobs — and that the bottleneck will be moving people into new roles fast enough. Reddit threads and policy discussions are already shifting from “will automation happen?” to “how do we train people at scale?”

The policy takeaway is practical: expect more investment in short, targeted reskilling (micro‑credentials, employer apprenticeships), and more pressure on community colleges and workforce agencies to adapt curriculums to hybrid human‑AI roles.

Source: Fortune on McKinsey’s analysis

Deep Dive

What a $30bn Aussie data centre IPO debacle actually means

Why this matters now: Firmus’s cancelled $30bn‑range IPO shows investor skepticism can rapidly puncture AI infrastructure valuations and slow capital flows for large, power‑hungry data‑centre projects.

Firmus Grid Ltd., a data‑centre operator promoted as a regional pillar for AI infrastructure and backed by Nvidia, abruptly pulled what would have been one of Australia’s biggest initial public offerings after investor demand evaporated. Banks and large U.S. fund managers that were expected to take big slices reportedly walked away in days, leaving the company no realistic path to the headline valuation it had been pitching.

“Firmus Grid Ltd. went from a rising star of the artificial intelligence data center boom to one of the biggest deal flops in recent memory,” as Bloomberg summarized the turn of events.

Why did this happen so fast? Several practical fault lines converged. First, the promise of AI‑driven demand doesn’t eliminate the brutal unit economics of data‑centre builds: land, grid connections, long lead times for planning and permitting, and massive upfront capex. Second, higher interest‑rate regimes and market volatility make lenders and institutional buyers pickier; future revenue projections that looked plausible in a frothier market look riskier today. Finally, investors often want named revenue streams or anchor customers before underwriting enormous valuations — and PR‑heavy narratives around “AI potential” are insufficient if the cash flows are unproven.

Investor rotation also matters: when markets decide to “name the winners,” flows tend to concentrate into obvious hardware and cloud names (think Nvidia, hyperscalers), leaving specialist operators vulnerable until they can demonstrate contracted revenue or long‑dated customer commitments. For the Australian market, the fallout may be broader: a failure on this scale could cool appetite for other infrastructure IPOs and force projects to seek private capital at more conservative terms, slowing regional data‑centre buildouts needed for on‑prem AI workloads.

Operationally, private fundraising will likely require Firmus to accept lower valuations or stricter covenants, and partners expecting a public exit will have to rethink timelines. The episode is a good reminder that shiny AI narratives still must clear mundane checks — grid access, power contracts, and balance‑sheet resilience — before investors will write the cheques the promoters want.

Source: FT coverage of the Firmus IPO pullback

Ukrainian drone strikes on Yandex datacenters knock cloud and payments offline

Why this matters now: Ukrainian strikes that damaged Yandex data centres show how fragile commercial cloud infrastructure can be in conflict zones, threatening consumer services and business continuity for customers hosted there.

Recent drone strikes reportedly hit multiple Yandex data centres outside Moscow, forcing outages across Yandex.Cloud services, mail, storage and parts of payments infrastructure. Yandex said operations at its Sasovo site were “completely halted,” and that in Kaluga “several modules have been completely knocked out,” forcing the company to reroute workloads and advise customers to use other zones as it assesses damage.

“Operations at our Sasovo facility were completely halted,” Yandex reported, noting fires and damage to facilities that host both commercial services and compute hardware.

This is a hard, physical reminder that cloud resilience is not merely software redundancy and multi‑AZ tricks; it depends on geography, power and transportation continuity. Data centres are dual‑use assets: they host consumer apps and payments but can also hold sensitive or governmental workloads. That raises legal and ethical complexity — strikes against facilities that support military or state functions risk being framed as legitimate targets by attackers, while defenders and customers see it as an attack on civilian infrastructure.

For businesses, the implications are immediate and practical:

  • Companies with critical services in the affected regions will need faster migration and contingency plans.
  • Global cloud providers and customers will rethink zone-level exposure: replicating to another country is costly but might be essential for business continuity.
  • Insurers and risk teams will reprice exposure where geopolitical conflict increases physical risk to hosted infrastructure.

Longer term, the strikes could accelerate investments in hardened, geographically distributed architectures and prompt customers to demand clearer contractual assurances about physical‑security measures and compensation for downtime. For operators, this is also a reputational challenge: beyond fixing hardware, they must rebuild trust that services will remain available and secure even as conflict dynamics change.

Source: The Register report on Yandex data centre strikes

Closing Thought

AI’s hype cycle is intersecting with two sober realities: capital markets are questioning big, promise‑heavy valuations, and the physical world still constrains cloud resiliency. That means startups and enterprises must marry narrative with hard contracts — power, customers, and contingency plans — if they want to survive the next stress test.

Sources