While the market spent early 2026 obsessing over core GPU constraints, a quieter shift was taking place at the edges of the ecosystem. I'm told that institutional focus is steadily expanding from foundational compute hardware toward the peripheral infrastructure needed to secure, power, and connect these massive AI workloads.

In this expanding narrative, Akamai Technologies (AKAM.US) is positioning itself as the security gatekeeper. Following its August 2026 launch of Workforce Protector for AI tools and a strategic alliance with Deloitte Canada to combat AI-driven cyber threats, the company is seeing a real shift in momentum. Its Q2 2026 total revenue of USD 1.1 billion grew a modest 5% year-over-year, but its Cloud Infrastructure Services segment surged an impressive 39% driven by AI workloads. This matters because it proves that AI can provide a legitimate second act for legacy cloud providers as traditional delivery networks mature.

And yet, further down the stack in physical hardware and communications, the reality is far messier. KULR Technology Group (KULR.US) boldly pivoted its messaging in June 2026 toward a "Physical AI" battery platform strategy targeting robotics and data centers. It even liquidated over USD 20 million in Bitcoin in July to pay down debt and fund non-dilutive growth. However, its Q2 2026 earnings revealed a significant miss on estimates with widening losses. The truth, as usual, is more complicated. Rebranding legacy thermal management tech into an AI play rarely translates immediately into clean financial metrics.

Similarly, Globalstar (GSAT.US) successfully launched 8 replacement satellites in August 2026 to bolster its direct-to-device communications network. Despite some recent institutional buying, the company reported a 3.5% year-over-year revenue decline for Q2 2026, alongside negative net margins and a pressured stock performance. Meanwhile, Magnite (MX.US), while peripheral to core AI hardware, is attempting to carve out its own niche by leveraging vast data processing capabilities in programmatic advertising, maintaining a delicate balance amid sector rotations.

On a more positive note, RingCentral (RNG.US) delivered a much cleaner execution in the cloud communications space. Its most recent 2026 quarter saw revenue hit USD 657 million, beating consensus and leading to an upward revision of its full-year EPS guidance. This fundamental strength, combined with a boosted quarterly dividend of USD 0.125 per share, drove the stock to a 52-week high in early September, easily outperforming many speculative infrastructure peers.

My view is that the margin for error in the broader AI and semiconductor sector has evaporated as we head into late 2026. The market is done rewarding vague promises of "AI enablement" and is demanding tangible revenue conversion instead. If you're still blindly buying into the periphery of the AI boom, good luck with that.

This article does not constitute investment advice.