AI startups surge, but privacy battles loom

Etched’s valuation jumped to $21 B after Jane Street deployed its first AI cluster system, prompting a fresh massive funding round. At the same time, Perplexity’s free AI offer in India added millions of users and lifted revenue by about 60 %.

Jane Street installed Etched’s shipped AI cluster and said the performance convinced the firm to lead the new round, according to the startup.^1234567 Perplexity saw its India revenue climb roughly 60 % after the Airtel‑sponsored free tier ended for new users, even as overall downloads fell, the report notes. Proton CEO Andy Yen, a longtime encryption advocate, warned that future AI systems will have to coexist with strong privacy guarantees.

Etched’s hardware play and soaring valuation

Etched delivered its first AI cluster to Jane Street earlier this month.²⁴ The deployment marked the company’s transition from software‑only services to a full‑stack hardware offering.¹²⁴ Jane Street’s engineering team ran benchmark workloads that matched or exceeded internal expectations, prompting the firm to write a check in the follow‑on round.^1234567

The funding round pushed Etched’s post‑money valuation to $21 B, a figure that represents a doubling of the company’s worth in just thirty days. The rapid increase reflects investor confidence in the market for purpose‑built AI infrastructure, a segment that rivals from the cloud giants are still scrambling to match.¹³⁴⁵⁶⁸

Analysts see the valuation spike as a signal that the hardware bottleneck is now a primary investment theme.¹³⁴⁵⁶ Companies that can ship turnkey clusters with low latency and high throughput are positioned to capture a share of the $200 B AI compute market that analysts project for the next three years. Etched’s move also pressures rivals like NVIDIA and AMD to accelerate their own on‑prem solutions.³⁵⁸⁷

Perplexity’s Indian user surge and revenue lift

Perplexity rolled out a free AI search tier in partnership with Airtel earlier in the year. The promotion attracted millions of new users in India, a market that accounts for a growing share of global AI consumption. When Airtel ended the offer for new sign‑ups, the company retained the bulk of the acquired audience.

Revenue from Indian operations rose about 60 % after the free tier closed, according to the source. The lift occurred despite a dip in overall app downloads, suggesting that the free offer succeeded in converting curiosity into paying usage rather than driving a fleeting download spike.

The pattern highlights a broader tension in AI product strategy: free access can seed a user base, but monetization hinges on retaining users once the incentive is withdrawn. Perplexity’s experience may prompt other AI startups to experiment with timed free tiers, especially in emerging markets where price sensitivity is high.

Proton’s privacy‑first AI warning

Proton’s chief executive Andy Yen has built the company around end‑to‑end encryption for email and VPN services. In a recent interview, Yen argued that AI models must be designed to coexist with strong privacy guarantees, even if the underlying algorithms are “un‑encryptable.”

Yen’s stance reflects a growing unease among privacy‑focused firms about the data‑intensive nature of large language models. Proton plans to integrate AI features that process user data locally, reducing the need to transmit raw inputs to cloud servers. The approach trades off the scale of centralized training for tighter user control.

Critics note that local processing can limit model size and accuracy, especially for complex tasks. Yen acknowledges the trade‑off but insists that privacy cannot be an afterthought. The debate underscores a split in the AI community between performance‑first roadmaps and privacy‑first architectures.

Industry context and what to watch

The three stories illustrate a converging pressure point for AI startups: hardware differentiation, user acquisition economics, and regulatory scrutiny over data handling.^1345687 Hardware providers like Etched are betting that bespoke clusters will outpace generic cloud instances.^1345687 At the same time, consumer‑facing AI apps are leveraging free tiers to build scale, as seen with Perplexity’s Indian rollout.

Regulators are tightening around data privacy, with the EU AI Act expected to impose strict transparency and risk‑assessment requirements. Proton’s privacy‑centric roadmap could become a template for compliance, but only if the company can deliver usable AI without sacrificing encryption.

Investors will watch Etched’s next hardware iteration for signs of performance gains that justify the $21 B valuation. Perplexity’s churn metrics after the Airtel promotion will reveal whether free‑to‑pay conversion can sustain long‑term revenue. Proton’s upcoming AI‑enabled VPN feature, slated for a Q4 beta, will test the market’s appetite for privacy‑first AI.

What to watch: the next funding round from Etched, Perplexity’s Q4 earnings release, and Proton’s beta launch of locally processed AI. Each data point will indicate whether the current growth strategies can survive heightened privacy regulation and the competitive push for faster, cheaper AI compute.

Updates

• 2026-09-14 — Jensen Huang puts Trump on speakerphone onstage to announce robots won’t take over the world (source)