Headcount Down, Compute Up

What happened
Block Inc., the parent company of Square and Cash App, said it will cut more than 4,000 jobs—almost half its workforce—as it restructures operations around AI. CEO Jack Dorsey described the move as necessary to embed AI across the company and said a smaller team equipped with AI tools can do far more than a larger pre‑AI workforce. News of the layoffs sent Block’s shares up around 25%.

Why it matters
Block is one of the first big consumer‑finance firms to slash headcount at this scale in order to pivot to AI‑first operations. The decision signals that AI efficiency gains are altering corporate labour budgets and could spur similar moves across industries. Investors rewarded the cost‑cutting plan, underscoring how financial markets are equally focused on AI adoption and profitability.

What's next
Dorsey suggested more companies may announce similar AI‑driven restructurings. Analysts expect the cuts to save Block $450–$500 million and predicted that competitors will be pressured to adopt AI to remain cost‑competitive. Watch for how Block reallocates resources into AI platforms and whether regulators scrutinise labour impacts.

What happened
A Reuters newsletter highlighted the rise of the “forward deployed engineer” (FDE) as AI’s hottest go‑to‑market job. These hybrid technologists embed with customers to integrate AI models into messy legacy systems. Anthropic’s legal and cybersecurity plug‑ins, released weeks ago, triggered steep sell‑offs in Thomson Reuters, RELX, CrowdStrike, Cloudflare, Okta and IBM stock underlining investor concern about disruption. Yet the article notes that demand for FDEs has grown 42‑fold, with only about 9,000 roles worldwide, and top labs are offering salaries up to $325k at OpenAI and $400k plus stock at Anthropic.

Why it matters
The FDE’s emergence shows that deploying AI isn’t just about models but about people who can translate breakthroughs into business value. With AI adoption linked to job losses in many sectors, this niche role commands premium pay because it blends technical fluency, domain expertise and client‑facing skills. It illustrates how AI may displace some jobs while creating specialised, well‑compensated ones that rely on tacit knowledge and judgement.

What's next
Expect AI labs and cloud providers to build teams of FDEs to accelerate enterprise adoption. The number of open roles is small, so demand could exceed supply for years. Over time, however, these jobs may morph into more automated integration tools or productised platforms, potentially reducing salaries as the function becomes standardised.

What happened
Meta Platforms signed a multi‑year, multi‑billion‑dollar deal to rent Google’s Tensor Processing Units (TPUs) to build new AI models. The arrangement supplements Meta’s existing chip suppliers: the social network has a $60 billion agreement with AMD and has also contracted with Nvidia for current and next‑generation chips. Google is pitching its TPUs as a cost‑efficient alternative to Nvidia GPUs and plans to open them to external customers via a joint venture.

Why it matters
The deal underscores how intense demand for AI compute is reshaping alliances among tech giants. By diversifying suppliers, Meta reduces its dependence on any single chip vendor and gains leverage over pricing. For Google, renting out TPUs not only monetises its hardware but also promotes an ecosystem that could challenge Nvidia’s dominance in AI training.

What's next
Meta will likely use the rented TPUs for training its family of large language and multimodal models. Google’s joint venture may open the door for other firms to access TPUs, increasing competition in the AI hardware market. Watch whether AMD and Nvidia respond with pricing changes or feature upgrades to retain customers.

What happened
Chip‑equipment maker ASML announced that its next‑generation extreme ultraviolet (EUV) lithography machine, known as the High‑NA EUV, has processed 500,000 wafers and is technically ready for high‑volume production. Each tool costs around $400 million and removes several manufacturing steps, enabling chipmakers like TSMC, Samsung and Intel to manufacture more powerful AI chips more efficiently. Integration into commercial fab lines could take two to three years.

Why it matters
High‑NA EUV represents the next leap in Moore’s law, allowing smaller features and higher transistor density. As current EUV tools approach their technical limits for AI chips, the new machines will be crucial for sustaining performance gains. Their steep price tag, however, means only top chipmakers can afford them, potentially widening the gap between the leading edge and the rest of the industry.

What's next
ASML will begin shipping High‑NA EUV tools to early customers in 2026. Chipmakers must redesign process flows and fabricate test wafers before full adoption. Investors will watch whether the tools meet yield and throughput targets, while governments may try to secure domestic access given the geopolitical importance of cutting‑edge lithography.

What happened
Anthropic CEO Dario Amodei said that his company cannot comply with a Pentagon request to relax safeguards that prevent its Claude models from autonomously targeting weapons or enabling domestic surveillance. The Pentagon warned it could remove Anthropic from its systems and label the lab a supply‑chain risk. Amodei said the contested contract, worth up to $200 million, never contemplated those capabilities. A Defense Department spokesperson insisted the Pentagon isn’t seeking autonomous weapons or mass surveillance but wants AI models available for all lawful purposes.

Why it matters
The dispute illustrates the tension between AI safety policies and military demand for advanced capabilities. Anthropic is taking a public stand to preserve its safety guardrails, underscoring the ethical complexities facing AI firms. The Pentagon’s threat to use the Defense Production Act or exclude Anthropic could set a precedent for government leverage over AI development.

What's next
Negotiations continue, with the Pentagon pressing for an answer by the end of the week. If Anthropic holds firm, the Department of Defense may turn to other labs like Google, xAI or OpenAI for classified contracts. The outcome could influence how AI companies set boundaries for national‑security work and how governments negotiate access to cutting‑edge models.

What happened
Dell Technologies expects revenue from its AI‑optimised servers will roughly double to about $50 billion by fiscal 2027 and noted that tech companies are expected to invest around $630 billion in AI infrastructure. Rising memory chip prices and supply constraints have pushed server prices higher, but Dell says those costs are being passed on to customers, helping protect margins. The company has sold AI servers to 4,000 customers and authorised a $10 billion share buyback while raising its dividend.

Why it matters
The forecast underscores how AI compute demand is translating into real revenue for hardware vendors. It also highlights the importance of memory chips and supply chains: constraints are inflating server prices yet improving profitability for suppliers. With hyperscalers spending at an unprecedented rate, infrastructure providers like Dell stand to capture outsized growth.

What's next
If memory chip prices remain elevated, server costs could continue to climb. Dell’s expansion suggests more competition among OEMs to supply AI infrastructure, and investors will scrutinise whether the projected doubling materialises. Look for updates on delivery lead times and potential partnerships with hyperscalers.

AI is no longer just a growth story — it’s a restructuring force. Companies are cutting headcount, bidding up elite talent, locking in compute alliances, and drawing ethical lines as intelligence becomes infrastructure and strategy at once.