I round up the most relevant AI-in-finance news - the deals being done, who’s rolling out what, and what’s actually working on the front lines
Vista Equity just became the first major PE firm to publicly commit to AI-driven job cuts…
…telling investors they'll reduce headcount by up to one-third. Salesforce paid $100 million for a startup that's been around less than a year. And $118 billion in AI infrastructure financing closed in the past month alone, structured like toll roads instead of tech deals.
Meanwhile, at the firms that rolled out enterprise AI firm-wide, the tools are failing at the simplest tasks.
In This Week’s Issue:
News Digest:
Vista Equity to cut workforce by up to 1/3 with AI
Salesforce acquires Doti AI for $100M
AI infrastructure financing hits historic scale
From The Trenches:
Why Copilot fails at the simplest PE tasks (and what actually works)
Other Cool Stuff I've Read or Seen:
AI bubble fears overstated, Anthropic disrupts cyber espionage, AGI predictions, and SoftBank's all-in bet
News Digest
Vista Equity to Cut Up to One-Third of Workforce with AI
Vista Equity Partners told investors it plans to cut up to a third of its 700-person workforce using AI, becoming the first major PE firm to publicly commit to AI-driven job reductions at scale. The $100 billion software-focused firm will automate roles compiling investor presentations, creating marketing materials, and aggregating deal data. CEO Robert Smith previously warned PE peers: "40% will have an AI agent, 60% will be looking for work." Vista scores portfolio companies on AI usage and is pushing them to build autonomous "AI agents."
The details:
CEO Robert Smith suggested staff could ultimately fall by as much as one-third through layoffs and reduced hiring
Targets: roles compiling investor presentations, creating marketing materials, and aggregating deal data
Vista directly employs 700 people; portfolio companies employ 10,000
Firm scores portfolio companies on AI usage and is pushing them to build "AI agents"
Smith previously told PE peers: "40% will have an AI agent, 60% will be looking for work"
Why it matters: First major PE firm to publicly commit to AI-driven workforce reductions at scale.
My take: They have to put their money where their mouth is. If your whole game is software investment and you've mandated all your portfolio companies become AI-native, you've got to walk the walk. Vista won't be the last. Every PE firm watching this will be running their own internal math on how AI changes their cost structure. The question isn't whether this spreads, it's how fast.
Salesforce Acquires Doti AI for $100M
Salesforce acquired Doti AI, an Israeli startup founded less than a year ago, for $100 million. The deal brings agentic AI tools for workplace automation into Slack's Israeli R&D hub. Doti focuses on AI agents that execute tasks with minimal human intervention. The acquisition is part of Salesforce's broader AI strategy, which includes pursuing an $8 billion deal for Informatica. The speed of the exit signals how quickly AI-native startups can command nine-figure valuations and how aggressively incumbents are moving to acquire capabilities before independent tools threaten core products.
The details:
Doti AI founded less than a year ago
Focus on AI agents that automate workplace tasks
Integration planned for Slack
Part of Salesforce's broader AI acquisition strategy (also pursuing $8Bn Informatica deal)
Why it matters: Shows incumbent tech racing to acquire early-stage AI capabilities before they become threats.
My take: You're going to see a lot more of this. Incumbents acquiring early-stage tech before it replaces them. Doti's been around for less than a year and commanded $100M. That's the new playbook: buy the disruptor before they disrupt you. Every established software company is looking at their core product and asking which AI-native startup could make it irrelevant. Better to write the check now than fight them in three years.
AI Infrastructure Financing Hits Historic Scale
Three massive deals in October-November mark a shift to project finance structures for AI infrastructure. Meta partnered with Blue Owl Capital on a $30 billion joint venture for its Louisiana data center, retaining just 20% equity. Oracle secured $38 billion in construction financing across 30+ banks for OpenAI's Stargate project. Anthropic announced a $50 billion investment with Fluidstack on November 12. All three use special purpose vehicles to keep debt off balance sheets, treating compute capacity like power plants rather than traditional tech capex.
The details:
Meta/Blue Owl - Hyperion (Louisiana): $30B joint venture, Meta retains 20% equity, Blue Owl owns 80%. $27B debt + $2.5B equity via SPV. 2GW capacity, completing 2029.
Oracle/OpenAI - Stargate: $38B construction financing across 30+ banks. Part of $300B five-year cloud deal. 4.5GW capacity across multiple US sites.
Anthropic: $50B US data center investment announced Nov 12 with Fluidstack. Texas and New York sites launching 2026. 800 permanent jobs, 2,400 construction jobs.
All using special purpose vehicles to finance infrastructure without direct balance sheet impact
Why it matters: AI infrastructure moving from tech capex to structured finance shows institutional capital betting on compute as an asset class.
My take: Look at the below GDP chart (insert below). 50% of GDP growth in Q2 2025 came from AI infrastructure and software buildout. This is the story of 2025. We're watching compute infrastructure get financed like power plants and toll roads. Meta doesn't want $27B of debt on its books, so Blue Owl writes the check and takes 80% ownership. That's not a tech deal. That's infrastructure finance. PE firms should be paying attention. The playbook for scaling AI isn't venture capital anymore. It's project finance, syndicated debt, and off-balance-sheet SPVs. The winners will be the firms that understand how to structure these deals, not just write software.
From The Trenches
Why Copilot Fails on the Simplest Tasks
A 50-person PE firm rolled out Microsoft Copilot firm-wide. It failed at their most basic, high-frequency task.
The firm reviews hundreds of CIMs annually. For each one, an associate reads the 90-page document and sends a firm-wide email with a five-sentence overview and key financials.
They tried: Feed the CIM into Copilot, ask for five sentences.
Result:
"Complete garbage." Not close. Not usable with edits. Unusable.
Why this fails: It's really hard to get consistent, repeatable results with LLMs. They're also not very good at large document context. What ChatGPT, Claude, and Copilot often do to save token costs is read the beginning and end of a document and make up the middle. You get spurious results.
The only way around this is to chunk up the information and feed it in systematically, only directing the model to the relevant pieces and cutting what's not.
Generic AI tools fail because they treat every document like a generic file. They don't know what matters in a CIM versus a legal brief. They don't have structured templates for what a PE firm actually needs. And they hallucinate when they hit token limits.
Other Cool Stuff I’ve Read of Seen This Week:
FT: AI bubble trouble talk is overblown (Nov 13) - Richard Waters argues supply shortage means no imminent disaster. Tell that to the LPs writing $30B checks for data centers.
Anthropic disrupts first AI-orchestrated cyber espionage campaign (Nov 13) - Chinese hackers used Claude Code to run cyberattacks autonomously across 30 targets. The robots are now hacking themselves. What could go wrong?
Verizon CEO: We'll hit AGI in next 2-4 years (Nov 12) - Dan Schulman thinks artificial general intelligence arrives by 2029. Either he knows something we don't, or someone needs to explain what "general" means.
Meta AI pioneer Yann LeCun discussed leaving to launch startup (Nov) - Chief AI scientist exploring departure after Meta cut 600 AI roles. Even the people building the robots worry about job security.
SoftBank dumps entire $5.8B Nvidia stake for OpenAI (Nov 11) - Sold all 32 million Nvidia shares to double down on OpenAI. Because nothing says "diversification" like going all-in on one unprofitable AI company
Acquisition Intelligence is a weekly newsletter on AI in M&A for finance professionals, private equity investors, investment bankers, corp dev teams, and deal-makers.
For questions, feedback, or to share what you're seeing in the market, reply to this email.
P.S. I'm Harry, co-founder of DealSage. We're building an AI-native M&A intelligence platform to help deal professionals turn their institutional knowledge into better decisions. If you're curious what we're up to, check out dealsage.io or just reply here