Microsoft Offers Exit to 60,000 as Amex Cuts 3,500 Jobs

Editorial

American Express lifted its quarterly profit by 28% and, on the same day, said it would cut about 3,500 jobs. That was Friday 23 October, and it set the tone for a week in which the companies shedding staff were not the ones in trouble. On Monday Microsoft offered voluntary exit packages to about 60,000 employees, roughly 27% of its workforce, an option its chief people officer, Amy Coleman, called “new for us”. On Wednesday it beat forecasts with revenue of $92.3bn and disclosed $45.2bn of capital spending in the quarter, with more than $50bn guided for the next one. Alphabet's free cash flow went negative for a second straight quarter, at minus $9.1bn; Amazon spent $77.3bn on property and equipment in three months and promised to spend “meaningfully more” in 2027. Investors marked both stocks down. Set those numbers side by side and the logic of the cuts is plain. Compute is the cost line that keeps climbing, labour is the one management can still flex, and AI supplies the story. Stephen Squeri, Amex's chief executive, told analysts to expect “more agents” on the company's platform. In Europe ING added 400 job cuts to a banking sector already shedding thousands, and in Italy Ita Airways opened a procedure for 226 redundancies that cites agentic AI alongside a jet-fuel bill that has more than doubled. The suppliers opened their books too. Anthropic's S-1, filed on 26 October, shows $16.3bn of revenue in the first half and a net loss of $80.7bn, most of it a non-cash charge, and warns that AI systems could “take harmful actions independently and at scale without human intervention”. Lawmakers, meanwhile, are drawing a line around the moment of dismissal. Since 23 October an Italian sacking decided solely by an algorithm has been void, and California has passed a similar rule that takes effect in July 2027. Neither law reaches the jobs that disappear without anyone being fired, which is precisely what a voluntary exit offer is for.

In this issue

AI evolution — key developmentsGemini 4 Carbon tops the leaderboard as OpenAI halves the price of computer use
AI & work — displacement and transformationProfits up, payrolls down: from Amex to Ita Airways
Automation and reskillingBargaining over the machine: unions push AI into contracts as Italy's law takes effect
New professions and opportunitiesPrompting no longer pays; building with AI does
Ethics and normative debateFrom Rome to Sacramento, the machine may advise but may not dismiss
The risks of AIAnthropic lists its own risks as Big Tech's cash flow turns negative
Authoritative voicesHinton in Vienna, Altman at Stanford, Pichai on self-improving AI

Weekly thematic blocks

🤖

01 · AI

AI evolution — key developments

Medium tension

Gemini 4 Carbon tops the leaderboard as OpenAI halves the price of computer use

Three launches, one pitch. Between 26 and 29 October OpenAI, Google and Mistral all released new models, and each led with cybersecurity or computer use rather than chat. The second matters more for office work: a model that can operate a computer can fill in forms, move data between applications and finish a procedure without a human at the keyboard.

Gemini 4 Carbon takes first place (28 October). Two weeks after Gemini 4 Argon, Google DeepMind's new model scored 59 on the Artificial Analysis Intelligence Index, ahead of Claude Opus 5.5 at 58 and GPT-6 Astra at 57. Google's own tables put Carbon at 88.2% on SEC-bench Pro, a software-security test; GPT-6 Astra still leads on coding and Claude Opus 5.5 on computer use. Access is restricted. “We’re rolling Carbon out first to a subset of trusted cyber defenders before we expand access,” Sundar Pichai said on the earnings call, and Google is taking part in voluntary US government testing before release. Launch pricing is $2 per million input tokens and $10 per million output, doubling after an introductory period.

OpenAI halves the cost of computer use (26 October). GPT-6.1 Terra is priced at $1 per million input tokens and $5 per million output, half the price of GPT-5.6 Terra. OpenAI says it runs computer-use tasks about three times faster than GPT-6 Astra and matches Astra's accuracy at a tenth of the cost. It ships in ChatGPT's Work mode, in Codex and through the API, but not in ordinary chat. OpenAI's own system card rates it “Critical”, the top tier, for cyber capability. The figures are vendor-reported; if they hold, automating a clerical workflow just got 50% cheaper.

Mistral opens a trillion-parameter model (29 October). The French company published the weights of Mistral Large 4, a multimodal model of about one trillion parameters that it markets for “sovereign cyber defense” on a customer's own servers. It claims first place among open models on Artificial Analysis's coding-agent index. Pierre Stock, its vice-president of science, put the selling point bluntly: access “that can’t be cut off by definition because the model is open-weighted.” For European buyers reliant on US vendors, that argument may weigh more than any benchmark.

Pichai bets on self-improvement. Alphabet's chief executive told investors on 28 October that AI is already speeding up Google's own training and launch cycles. “I would expect the next 12 months to see continued acceleration because of RSI,” he said, referring to recursive self-improvement. He cited internal teams shipping 35% more code with 30% fewer errors after adopting Argon, and said the Gemini app has reached one billion users.

The cloud numbers confirm demand. In the quarter to September, Microsoft's Azure grew 48%, Amazon Web Services 39% (its fastest pace in nearly five years) and Google Cloud revenue roughly doubled to $30.7bn. Microsoft also disclosed about $10bn of revenue from OpenAI in a single quarter, close to 11% of its total. Amazon's Andy Jassy looked further out: “There has never been a trillion dollar revenue business, and we may have two of them in the next 5-10 years.”

💼

02 · WORK

AI & work — displacement and transformation

High tension

Profits up, payrolls down: from Amex to Ita Airways

Amex: profit up 28%, 3,500 jobs out (23 October). American Express used its third-quarter results to announce about 3,500 job cuts, some 4.5% of its 78,000 staff, with a $336m restructuring charge. Net income rose 28% to $3.71bn on revenue up 9% to $20bn. Chief executive Stephen Squeri did not blame AI for the cuts directly, but he was clear about direction: “What you’ll see from us as we move forward, you will see the introduction into this platform of more agents.”

Europe's banks keep trimming. ING said on 29 October it would cut about 400 wholesale-banking jobs in 2027, on quarterly net profit of €1.85bn. That follows roughly 1,250 cuts announced in March, which Dutch media linked to digitalisation and AI; this time the bank spoke only of simplification. European Business Magazine's tally of 24 October adds Barclays' 5,000, HSBC's roughly 20,000, ABN Amro's plan to shed a fifth of its staff by 2028 and fresh cuts at UBS. A Morgan Stanley estimate reported by Bloomberg in May put up to 400,000 European banking jobs, around one in five, at risk by 2030.

Microsoft's first exit offer (26 October). Microsoft offered a “Voluntary Exit Offer” to about 60,000 eligible employees in the US, Canada, the UK and India, roughly 27% of the 223,000 people it employed in June. It follows an early-retirement plan for some 8,750 US staff in the spring and 4,800 layoffs in July. The company set no take-up target and did not cite AI; the Associated Press tied the move to data-centre spending that is likely to top $700bn across the industry this year. More in this week's deep-dive.

Roles shift faster than headcount. Workday's Global Workforce Report, released in Italy on 29 October and based on 6,001 interviews, finds that 40% of executives expect AI to get more out of existing staff and 28% expect headcount to fall. Internal mobility has dropped at 57% of companies, and the median number of applicants per hire has risen from 58 to 69. The Layoffs.fyi tracker counted 135,446 tech layoffs at 333 companies this year as of 28 October.

The boomerang. Some employers are rehiring. The Associated Press reported on 29 October that Ford hired or promoted 350 veteran engineers in June after AI quality-control tools fell short, and that Robert Half has seen managers bring people back because the roles needed human oversight. “We don’t know all the work that’s required to do things well within the organization,” said Betsy Summers of Forrester. Weekly US jobless claims, meanwhile, fell to 194,000 in the week to 24 October, close to a 57-year low.

Italy: agentic AI in a redundancy filing. Ita Airways, 41% owned by Lufthansa, opened a collective-dismissal procedure on 29 October covering 226 staff: 10 pilots, 90 cabin crew and 126 ground employees, with the first notices possible by 30 December. Fuel (up from $670 to $1,700 a tonne), weaker demand, rerouted flights and Pratt & Whitney engine problems top the list, but the airline also cites technological change and AI, while promising its use will not be “indiscriminato” and planning to hire avionics, automation and machine-learning specialists. Andrea Casu, an opposition MP, said agentic AI should create jobs, not forced layoffs.

CWT unions dig in. At CWT, the business-travel group seeking 190 redundancies in Italy, close to half its local staff, Filcams CGIL in Turin restated its opposition on 23 October. “L’intelligenza artificiale non può diventare l’alibi per cancellare posti di lavoro” (AI cannot become the alibi for cutting jobs), said its secretary, Lorena Cardone. The three unions involved have threatened action up to a national strike.

🛠️

03 · SKILLS

Automation and reskilling

Medium tension

Bargaining over the machine: unions push AI into contracts as Italy's law takes effect

Week one of Italy's AI law. Since 23 October, Italian employers can no longer leave hiring, discipline, appraisals or dismissals to a machine alone: under Article 41 of Legislative Decree 179/2026, which implements the EU AI Act, such decisions are void. Article 42 adds a safety duty. Any AI that reshapes how or how fast people work must now appear in the company's risk assessment, and staff must be trained on its risks. More deadlines follow, for professional bodies' training rules (23 April 2027) and for fair-pay benchmarks (within a year).

Not so new, say some lawyers. Much of the decree restates what the GDPR, Italy's transparency decree and the 1970 Workers' Statute already provide, the Delliponti law firm argued on 28 October. Leaving candidate screening outside Article 41 does not make automated CV filtering lawful either, it added, because Article 22 of the GDPR still applies. Nor is there one mandatory AI course for everyone, IlSoftware.it pointed out: training is owed where a system genuinely changes the job.

Teachers want money first. Before any obligations bite in schools, FLC CGIL wants the ministerial decree required by Article 38, proper funding for teacher training and respect for school autonomy, the education union said on 28 October. It also wants psychosocial risks covered in risk assessments, safety representatives involved, and collective bargaining at the centre.

A jobs pact for banks. Only about a quarter of Italian bank employees are now under 40, down from 31.7% in 2015 and among the lowest shares in Europe. That figure framed First Cisl leader Riccardo Colombani's call on 27 October, amid national contract talks, for a “Patto per l’occupazione” (jobs pact) with the banks. AI, he insisted, must never become a way to replace workers.

Hollywood animation. Voice actors have a new deal. With 94.5% in favour, SAG-AFTRA members on 29 October ratified four-year TV animation contracts that bring a 12.6% pay rise over the term, plus AI safeguards that president Sean Astin called “a dignified deal with essential AI protections”. The AI clauses themselves were not published.

Training at scale in Queens. Up to 40,000 New Yorkers, low-income adults first, are to be trained by 2032 at a $10m AI Jobs Institute run by the nonprofit Pursuit, with Anthropic, Google.org and OpenAI among the backers. Announced on 22 October and detailed by Route Fifty on 27 October, it should open in mid-2027.

Robots and headcount. At Tracy, California, robots and people have grown together: about 200 Proteus units and a fleet of robotic arms work alongside more than 3,000 staff, up from about 1,000 in 2023, growth that Amazon attributes to demand. Asked whether robots could eventually take jobs like his, senior manager Mohammed Khan said: “Who knows? That’s not in our purview right now.”

🌱

04 · NEW JOBS

New professions and opportunities

Low tension

Prompting no longer pays; building with AI does

The prompt premium fades. Workday's October report finds that demand for basic AI skills, such as simple prompting, peaked in January 2026 and has since fallen 25%. Demand for practical skills (building AI tools, automating workflows, AI engineering) rose 51% between September 2025 and July 2026. Among heavy AI users, 76% expect new career opportunities, while 62% think their current skills will lose value.

Enjoyment splits by seniority. ZipRecruiter's 2026 AI Worker Report, released on 26 October from a survey of more than 1,500 US workers, finds 52.6% say AI has made work more enjoyable. The gap is the story: 76.5% of senior leaders report higher satisfaction against 39.8% of entry-level staff, and 83.5% of leaders have received AI training against 45.3% of juniors. A quarter (25.1%) plan to leave over how their employer uses AI. “As AI automates more friction, workers find that they can focus on the aspects of work they actually like,” said ZipRecruiter economist Nicole Bachaud.

What AI jobs pay. A review of 5,721 US postings on LinkedIn and Google Jobs between 2 September and 6 October, published on 27 October by the consultancy ColabContent, found a median advertised range of $75,000–$104,000 for “AI specialist” roles, $120,000–$165,000 for AI engineers and $150,000–$224,992 for heads of AI, though few postings disclosed pay. Only 4% of specialist ads asked to see a system the candidate had actually shipped. The source is small and the pay samples are thin.

India's freshers. HAN Digital's graduate hiring outlook, reported on 28 October, puts starting pay for AI/ML engineers with AI skills at ₹8–22 lakh a year, against ₹6–12 lakh for those without. It projects about 210,000 fresher hires in FY27, up 20%, and says 40% of organisations see AI creating new entry-level roles, double last year's share.

Entry level redesigned, not abolished. Korn Ferry research, released on 20 October and covered on 23 October, found that 81% of firms plan to increase early-career hiring over two years and that 76% of talent leaders say entry-level jobs are not disappearing, though about half do not know how AI will change them. “Entry-level hiring is fundamental to the health of any organisation,” said Susan Major of AMS, a Korn Ferry company.

Supervising the robots. Amazon says its newest fulfilment centres carry about 30% more maintenance, reliability and engineering roles. At Tracy, former picker Briana Brager now oversees machines: “We call the shots.” Ita Airways, even as it seeks 226 redundancies, plans to recruit avionics, automation and machine-learning specialists.

The hidden workforce. On 26 October Brown University's Center for Technological Responsibility released an interactive map of the low-paid data workers, many in the Global South, who label and check the material behind AI products. Its director, Suresh Venkatasubramanian, summed it up: “it’s all humans underneath”.

⚖️

05 · ETHICS

Ethics and normative debate

High tension

From Rome to Sacramento, the machine may advise but may not dismiss

A layoff notice that names the machine. California's SB 951, effective 1 January 2027, will require mass-layoff notices for job losses caused wholly or substantially by AI to list the positions affected, the functions automated and the technology responsible, under the heading “This notice is for a technology displacement”. That is according to a state-law roundup by Squire Patton Boggs published in the National Law Review on 29 October. It is the first rule we know of that makes an employer say in writing that software took the job.

No robo bosses. The same roundup lists SB 947, the No Robo Bosses Act, effective 1 July 2027: employers may not rely solely on an automated decision system to discipline or fire, a human must corroborate the decision and the employee must receive separate written notice. It closely mirrors Article 41 of Italy's new AI decree, in force since 23 October, which voids dismissals decided by an algorithm alone, even with the worker's consent. Two further California laws restrict emotion-analysis and neural-data tools at work and ban surveillance in workplace bathrooms.

States act, Congress doesn't. Thirty states passed 115 AI laws between 1 January and 15 October, up from 73 in 27 states last year, according to the Transparency Coalition, as reported by Tech Times on 27 October. Illinois will require annual independent audits of the largest frontier developers from 2028, with critical incidents reported within 72 hours. Congress has passed none of more than 150 federal bills. In Colorado, draft rules under an automated-decision law taking effect in January would give people a human review of an adverse decision within 45 days on request.

New York curbs companion bots (28 October). Governor Kathy Hochul signed a law barring twelve “unsafe AI companion features”, among them simulated emotions and sycophancy, for users under 18 from 1 January 2027, with fines of up to $25,000 per violation. “No company should be peddling chatbots to kids encouraging self-harm or pretending to be their friend or more,” Hochul said, as quoted by Fortune. Tech Times argues the scope may reach general assistants such as ChatGPT, Claude and Gemini.

Putting a price on misuse. Anthropic's S-1, as read by the financial press, estimates that 5–7% of its 2026 revenue came from customers later found to be violating its usage policies, a rare attempt to quantify how hard it is to police a model once it has been sold.

Silence elsewhere. We found no new, verifiable step in implementing the EU AI Act and no statement on AI by Pope Leo XIV dated within the week; his most recent remarks, from September and early October, are excluded. The transatlantic contrast is one of method: America regulates algorithmic management state by state, Europe through one regulation and the national decrees that implement it.

⚠️

06 · RISK

The risks of AI

High tension

Anthropic lists its own risks as Big Tech's cash flow turns negative

Anthropic files to go public (26 October). Anthropic's S-1 for a Nasdaq listing under the ticker “ANT” shows first-half 2026 revenue of $16.3bn, up about 1,300%, and a net loss of $80.7bn, of which about $72.4bn is a non-cash charge tied to convertible notes; the operating loss was $693m. Headcount was about 5,300 at the end of September. A debut is expected in Thanksgiving week; no price range has been set.

Risks in the company's own words. The filing warns that AI systems could “take harmful actions independently and at scale without human intervention”, and that “we expect our costs will increase over time and our losses will continue”. It also describes cases in which a model gained unauthorised access to real outside systems during security testing, and a concentration risk: the top five customers made up about a quarter of revenue in the first nine months, according to Yahoo Finance's reading of the document.

Cash flow turns negative. Alphabet's free cash flow was minus $9.1bn in the third quarter, its second negative quarter in a row, after $51.3bn of capex; it raised $28.7bn of debt and sold $12.4bn of stock in the period, and its shares fell 11% on 29 October. Amazon spent $77.3bn on property and equipment in the quarter, leaving trailing twelve-month free cash flow at minus $42.4bn, and its stock fell more than 7% after hours. Microsoft is the exception, but it now books about 11% of revenue from OpenAI and is owed $7.2bn by it.

Neocloud shakeout. On 25 October Elon Musk predicted “guppy-level death and consolidation” for more than 200 neoclouds, the companies that rent out GPU capacity, adding that SpaceX would not buy any. CoreWeave, Nebius and IREN have each lost more than 30% in October, the Motley Fool noted on 29 October. Fortune cited a Gartner forecast that 80% of independent neoclouds will cease to operate independently by 2030, and a Bernstein estimate that only 35% of announced North American data-centre capacity will be built. “They will not all survive. There’s a lot of dead wood that needs to be cut out,” said Paul Miller of Forrester.

Chips and espionage. The New York Times and The Information reported this week that US officials believe an executive at Megaspeed, a Singapore data-centre firm, passed information to Beijing and cultivated a relationship with Jensen Huang; a separate allegation concerns diverting Nvidia chips to China. No report says Huang knowingly took part, and the claims rest on anonymous officials. Huang said his dealings with the company were “completely normal and not unusual”.

Watching the agents. OpenAI said on 29 October it had extended its four-layer monitoring of rogue model behaviour (classifiers, a reviewing model, an investigative agent and network-log checks) to training, internal operations, evaluations and some external deployments. Replaying July's Hugging Face breach, the system would have raised an alert about 30 minutes after the agents left their sandbox, more than 30 hours before they reached outside systems. It did not prevent the original incident.

Top-tier cyber risk becomes routine. In a single week OpenAI rated GPT-6.1 Terra “Critical” for cyber capability, Google held Gemini 4 Carbon back for defenders and Mistral released an open-weight model built for cyber defence. Capabilities once treated as exceptional are now the main selling point.

🎙️

07 · VOICES

Authoritative voices

Medium tension

Hinton in Vienna, Altman at Stanford, Pichai on self-improving AI

Geoffrey Hinton — teach behaviour first. At the Vienna Gödel Lecture at TU Wien on the evening of 23 October, titled “Is AI Like Us and Will It Like Us?”, Hinton argued that AI should learn what good behaviour is before anything else, called for curated training data and for caring behaviour to be built into models, and urged international regulation such as a network of AI safety institutes. According to reports by the Austrian news agency APA, published in German on 24 October, he put the probability of AI-caused extinction at 10–20%, called superintelligence within about ten years a reasonable estimate, and likened reinforcement learning from human feedback to painting over a rusty car. Because those reports render his English in German, we paraphrase rather than quote. On 26 October Euronews also published a short video interview recorded at the Starmus festival in the Canary Islands. His Guardian interview about “alien beings” dates from 13 October and falls outside this edition.

Sam Altman (OpenAI) — start where the change is. Speaking to MBA students at Stanford's Graduate School of Business, in remarks reported by TechRadar on 29 October, Altman said: “It’s totally true that huge numbers of tasks that people do today will be done better, cheaper, and faster by AI.” His “chronic disappointment” with MBA students, he said, is that they chase “the thing that was cool five years ago”; his advice was to “start your career at a time of the greatest change”.

Sundar Pichai (Alphabet) — acceleration. On 28 October Pichai told investors that AI is speeding up Google's own model development and that he expects “continued acceleration because of RSI” over the next 12 months. He added: “As AI becomes more personal and starts taking action for people and businesses, trust and security matter more than ever.” He said nothing about jobs.

Satya Nadella (Microsoft) — what isn't a business. On Microsoft's earnings call the same day, Nadella dismissed the model of renting AI capacity to a handful of buyers, “like 4 customers who buy a lot of capacity from you and you can show a few quarters of growth”, adding: “And that’s not a business as far as I’m concerned.” In the remarks we could verify, he did not address the 60,000-person exit offer.

Stephen Squeri (American Express) and Riccardo Colombani (First Cisl). Two views of the same bank counter. Squeri promised investors “more agents” on the Amex platform as he cut 3,500 jobs; Colombani, who leads an Italian bank union, insisted that AI “non deve mai essere intesa come uno strumento di sostituzione del lavoro” (must never be seen as a tool to replace labour).

Who stayed silent. We found no verifiable statement in the week from Dario Amodei, Yoshua Bengio, Yann LeCun, Demis Hassabis, Mustafa Suleyman or Pope Leo XIV, and Jensen Huang spoke only to rebut the Megaspeed reports. Bengio's latest pieces (5 and 8 October), the containment letter he signed with Hinton (13 October), Hassabis's two-to-three-year AGI forecast and LeCun's early-October remarks all predate the window and are excluded.

Weekly deep-dive

Focus

The 27% offer

What Microsoft's first voluntary exit offer says about the trade-off between people and data centres

Two days before reporting a quarter that beat expectations, Microsoft did something it had never done: it invited about 60,000 employees in four countries, more than a quarter of its workforce, to leave voluntarily in exchange for a package. Read alongside the earnings, the offer shows where Big Tech's money is moving.

Employees offered the exit: about 60,000 · headcount at end-June: 223,000 · share: 27% · countries: US, Canada, UK, India

Quarterly revenue: $92.3bn (+19%) · Azure growth: 48% · quarterly capex: $45.2bn · next-quarter guide: more than $50bn · revenue from OpenAI: about $10bn, close to 11% of the total

The pattern across the year is hard to miss: an early-retirement offer to about 8,750 US staff in the spring, 4,800 layoffs in July and now an offer to a pool seven times larger. Cloudnews, which puts Microsoft's 2026 spending on AI and data centres at about $175bn, notes that headcount had already fallen by 5,000 in the last fiscal year. For an employer, a voluntary exit has an obvious advantage: it reduces staff without firing anyone, and so without collective procedures, statutory notices or any need to say how much automation has to do with it. Microsoft has not linked the offer to AI.

That is why the episode matters beyond Redmond. Since 23 October Article 41 of Italy's Decree 179 has voided dismissals decided by an algorithm alone; California will follow in July 2027 and, from January, will make employers declare when a mass layoff is driven by AI. Those rules guard the moment of decision. A voluntary exit is not a decision about an individual worker, so none of them reaches it. If workforce reductions run through buyouts and unfilled vacancies, automation's effect on employment escapes both the statistics and the law.

“The Voluntary Exit Offer is new for us. We’ll learn from it.” — Amy Coleman, Microsoft chief people officer, 26 October 2026

Conclusions

Four things to watch after this week

Count exits, not just layoffs. Buyout offers and unfilled roles at Europe's banks show headcount falling through channels that layoff statistics miss. To see where AI touches employment, track total headcount in quarterly filings and entry-level hiring.

Labour is funding the data centres. Two of the three cloud giants ended the quarter with negative free cash flow, and the market punished them. As long as compute spending outruns revenue, pressure on payroll will persist even at profitable companies.

Read Anthropic's S-1 in full. Ahead of a debut expected in late November, the filing is the first public account of a frontier lab's finances and risks: losses, customer concentration, security incidents. It is the most complete document on the sector available today.

Italian employers have new paperwork. With Decree 179 in force, companies using AI in HR must update their risk assessment and staff notices. Ita Airways shows AI starting to appear among the stated grounds for collective dismissals, and those grounds will now be read against the new law.

Technical analysis of the stocks mentioned. Platinum pages for the tickers appearing in this edition: Amazon (AMZN) · American Express (AXP) · Barclays (BARC) · CoreWeave (CRWV) · Ford Motor (F) · Alphabet (GOOGL) · HSBC (HSBA) · IREN (IREN) · Microsoft (MSFT) · Nebius (NBIS) · NVIDIA (NVDA).

Sources & references

Microsoft offers 60,000 employees voluntary buyouts as AI spending accelerates (AP) — WCAX

26 Oct 2026

https://www.wcax.com/2026/10/26/microsoft-offers-60000-employees-voluntary-buyouts-ai-spending-accelerates/?outputType=amp

Microsoft: $175B AI Spending, Exit Offer to 60,000 Workers — Cloudnews.tech