Companies Are Spending 5x More on AI Now
Enterprise AI spend rose nearly 5× per company in 2026, largely because agents shift usage from per-seat subscriptions to token-intensive work.
- Plan AI budgets around usage and tokens, not just employee seats; companies spent about 2.5× what they had budgeted.
- Agentic coding led application growth; coding-agent revenue rose from $1B to $24B, while other coding platforms also grew.
- Most new AI spending came from net-new budgets, not existing software budgets, though agent-driven software disruption remains a risk.
The report’s figures and implications are summarized here; the episode adds commentary, not a detailed demo. · Best for: Business leaders, finance teams, and AI or software market analysts
The episode reviews enterprise AI adoption and a survey of about 400 US enterprise decision-makers. Its central argument is that agents drove a sharp increase in spending per company, while adoption and market share varied by model, application category, and incumbent status.
Market headlines and product updates1:49
- The host reports a broad market sell-off and uncertainty about Anthropic’s potential IPO.
- An independent safety report said Anthropic’s model improvement pace roughly doubled over the past year, attributing the acceleration mainly to AI assistance in research.
- Product updates included Claude personal-finance access through Plaid, voice and video for Meta’s Muse agent, and animated GIF support in OpenAI’s image model.
Enterprise spending accelerates13:59
- The survey estimates enterprise AI spending rose from $33B in 2025 to nearly $177B in 2026.
- The share of enterprises using AI rose from 55% to 65%, while average spend per AI-using enterprise increased 4.7×.
- Companies spent about 2.5× their planned amount; the host says budgets need to account for token-driven usage and agents, not just seats.
Model choices and open-source usage19:30
- Spending shifted toward frontier models: their share rose from 35% to 63%, despite price competition among smaller models.
- Anthropic’s share of enterprise API spend reached 55%; OpenAI recovered to 31%, while Google fell to 9%.
- Open-source models represented 19% of tokens but 4% of spend; Kimi accounted for 61% of open-source spending.
Applications and agentic coding24:58
- Revenue for coding agents from the labs, including Codex and Claude Code, rose from $1B to $24B; other coding platforms grew from $2.6B to $7.8B.
- Financial-services apps grew 8× and agentic-coding apps 7.7×; marketing grew 1.8×.
- AI-native companies led coding, finance and operations, and design; incumbents retained more than half of spend in several other enterprise categories.
Software budgets and disruption28:48
- The survey found the SaaS application market grew 11% in 2026, down from 13% in 2025.
- Of accounted-for AI spending, 55% came from net-new budgets and 21% from existing software budgets.
- The host cautions that agents using software, including headless software, could still disrupt SaaS over time.
Step by step
- When forecasting AI spend, separate per-seat adoption from token- and agent-driven usage.
- Track usage and spending by model and application category; compare actual usage with budget assumptions.
- Review whether AI tools add new capabilities or replace existing software, and update procurement plans accordingly.
Tools & resources mentioned
- Menlo’s State of Enterprise AI report — survey data and spending estimates discussed in the episode.
- Plaid — provides read-only bank-account connections for Claude’s personal-finance feature.
- Claude — discussed for personal finance and as Claude Code, a coding agent.
- Codex — OpenAI coding agent included in the coding-agent revenue discussion.
- Cursor, Cognition, Factory, and Lovable — other coding platforms whose revenue growth was discussed.
- Kimi — identified as the largest share of open-source model spending in the survey.
Notes are AI-generated from the video's transcript and may miss nuance.