VENTURE CAPITAL
VENTURE LENS
Early R&D can be variable by design. Production costs need a traceable relationship to customer usage, pricing, and value delivered.
DILIGENCE
Ask for evidence
Can the company show cost per customer, task, or workflow as a distribution, with the method and assumptions visible?
COMPANY BUILDING
Protect the learning loop
Keep experimentation visible without mistaking it for recurring product cost. Identify production workloads and their owners.
PORTFOLIO SUPPORT
Protect the learning loop
Help portfolio teams maintain an evidence pack for margin trajectory, provider costs, and measured product outcomes.
EVIDENCE FROM THE BENCHMARK DATABASE
46.9% of enterprises ran over budget on AI
· The Futurum Group
46.9% of enterprises reported running over budget on AI, against 31.8% approximately in line with budget and only 5.6% below plan
Only 17.2% of over-budget organizations pause or reduce the AI initiative
· Techstrong.ai
Among enterprises over budget on AI, only 17.2% pause or reduce the AI initiative itself, while 48% request additional budget and 43% absorb the overrun and settle it later
61% cut external contractors and consultants to fund AI overruns
· Techstrong.ai
Organizations over budget on AI fund the gap by cutting elsewhere: 61% reduce external contractors and consultants, 60% cut IT support and helpdesk, 58% cut software and application development, and 23% reallocate from a non-IT business unit budget
10% have no formal AI budget and 5.6% do not know where their AI spend stands
· The Futurum Group
10% of enterprises have no formal AI budget to measure spending against and a further 5.6% do not know where their AI spending stands, meaning roughly one in six cannot say whether they are over or under budget
US$795 billion of 2026 hyperscaler capex rising to nearly US$1.08 trillion in 2027
· Reuters (via Investing.com)
Microsoft, Alphabet, Amazon, Meta Platforms and Oracle are expected to spend a combined US$795 billion on capital expenditure in 2026, rising to nearly US$1.08 trillion in 2027, against total AI spending of nearly US$800 billion in 2026
42% would expand existing AI workloads if token prices fell 25% or more
· CIO Dive
42% of surveyed executives said they would expand existing AI workloads if token prices fell 25% or more, while only about 10% of workloads need advanced frontier-level reasoning
PORTFOLIO QUESTIONS
Make the next conversation specific.
[ 01 ]
What is AI cost per customer or agent run, and how does it vary across usage?
[ 02 ]
Which spend is experimentation, and which supports a sold product?
[ 03 ]
What evidence connects product AI cost to retention, revenue, or another measured outcome?
[ 04 ]
Does tracked usage reconcile to provider invoices?
[ 05 ]
How do unit costs change under the next planned scale scenario?
