5G takes the lead, the hiring crisis deepens, and rivals team up

April 1, 2026

Happy April Fools!

Private 5G deployments have overtaken LTE for the first time, the data center industry is running out of people to hire, and an AI compliance startup is under investigation for allegedly fabricating audit evidence.

As always, thank you for reading - and hit us up with any feedback.

Today’s edition:

Let’s dive in.

1. According to Uptime Intelligence research reported by Network World, nearly half of data center operators struggle to find qualified candidates and 37% can't retain the staff they already have, largely because the industry defaults to poaching talent from competitors rather than building it internally. With operations management roles now overtaking junior positions as the primary shortage area, the "silver tsunami" of retirements is arriving faster than pipelines can replace it.

2. Fierce Networks published a useful catchup on the orbital data center space, and the timing is relevant: Starcloud just raised $170 million at a $1.1 billion valuation, adding momentum to a category that now includes SpaceX, Google, and Axiom Space. The core pitch across all of them is the same - sun-synchronous orbits provide near-continuous solar power at a fraction of terrestrial energy costs, which looks increasingly attractive as land and power constraints slow down ground-based data center builds.

3. Cisco and HPE filed a joint comment with the FCC asking regulators to extend 6 GHz unlicensed spectrum access to cruise ships, arguing that the original exclusion was based on interference concerns that no longer apply. The two companies contend that cruise ship hulls actually attenuate signals better than most land-based buildings, making the legacy restriction unnecessary.

4. An anonymous Substack account published allegations that AI compliance startup Delve fabricated audit evidence and misled over 1,000 clients in 50 countries into believing their companies were compliant, prompting the company to halt all product demonstrations while it investigates. The broader takeaway for enterprise buyers is straightforward: compliance claims from AI vendors need independent verification, and a polished demo is not a substitute for one.

5. The FCC proposed capping the percentage of customer service calls telcos can route to foreign call centers, requiring overseas agents to speak proficient American English, and mandating disclosure when calls are handled outside the US, though labor union CWA says the rules miss the bigger threat, which is AI replacing call center jobs entirely.

🇿🇦 Equinix plans to build more facilities in South Africa as part of a $438 million plan to capitalize on the continent’s artificial-intelligence boom. (Moneyweb)

💰 Nile launched what it's calling Nile 2.0, adding native zero trust capabilities directly into its NaaS platform (The Fast Mode)

💤 A new report published by Rapid7 Labs finds evidence of ‘Chinese sleeper-cell’ backdoors hidden in critical US telecommunications equipment. (Rapid7)

🌐 Arm expands its compute platform to silicon products with the introduction of a new data-center CPU. (Arm)

🍎 Apple continues crackdown on ‘vibe coding’ apps with removal of additional AI tools from App Store under Guideline 2.5.2 (Mac Rumors)

🕷️ OpenAI patches ChatGPT flaw that smuggled data over DNS. (The Register)

The private wireless market crossed 2,000 enterprise customers tracked by GSA member companies at the end of 2025, and 5G has now overtaken 4G LTE in new deployments for the first time. Manufacturing remains the largest segment by deployment count, with education and mining also prominent. Most organizations are running more than one private network, and the average deployment is generating over $115,000 in annual revenue for the vendor ecosystem.

Between the lines: The 51% LTE share of the installed base tells you something important: this market is not a clean 5G transition story. It's two parallel tracks. LTE-based systems are still being deployed where economics and maturity favor them, while 5G is winning where capacity, latency, or device roadmap requirements demand it. For enterprise network architects, the practical implication is that 5G SA is increasingly the right call for new builds, but LTE isn't going away as a maintenance and expansion platform for existing deployments anytime soon.

The broader read: Spectrum policy is quietly becoming a competitive differentiator. GSA's data shows that markets with clear, enterprise-friendly dedicated spectrum allocations are moving from experimentation to industrial-scale deployment, while fragmented spectrum environments are pushing buyers toward managed or hybrid models. As the market expands to 84 countries, how regulators handle spectrum will increasingly determine which vendors and integrators win in which geographies.

Download the full report via Global Mobile Suppliers Association

AI inference is getting cheaper. Your total bill probably isn't.

Gartner forecasts that running inference on a trillion-parameter model will cost providers over 90% less by 2030 than it does today, driven by better chips, higher utilization, inference-specialized silicon, and edge processing. On paper, that's a dramatic efficiency gain.

The catch is on the consumption side. Agentic models require between 5 and 30 times more tokens per task than a standard chatbot. Token costs fall, token demand rises faster, and the net result for enterprise buyers is that overall inference spend goes up, not down. Providers aren't expected to pass the savings through in full either, partly because the largest AI labs still aren't profitable.

What this actually points to is a split market. Routine, high-volume AI tasks trend toward commodity pricing and get absorbed into platforms or undercut by open source. Frontier reasoning, complex agents, and high-stakes automation stay expensive and remain gated behind the use cases that can justify the cost.

For IT leaders evaluating AI infrastructure spend, the number to watch isn't the per-token rate. It's total token consumption as agentic workloads scale. Cheap tokens and a rising bill aren't exactly a contraction here, but more than likely the plan.

Read more on LLM inferencing predictions here.