The CFO's statement about allocating funds for compute capacity is a general bucket, like saying I'll buy 'groceries' for the week; it's a 2 out of 10 for specificity. Mistral AI's plan for 200 MW by 2027 and 1 GW by 2030 is far more detailed, probably an 8 on that same scale, like knowing I need 2 pounds of organic carrots and a gallon of 2% milk. A budget category does not automatically translate to a project specification with power consumption targets. For example, my budget says 'new server hardware,' but that doesn't tell IT if it's a blade server or a rackmount, or the exact power draw; it's a ratio of commitment, not an absolute. More specific engineering plans and demand forecasts would be needed to bridge that 6-point gap in detail.