In brief. The SoftBank case shows the cost of changing a decision made to secure supply. An investment committee needs to know the price of the plan and the price of revising it. In its 2026 half-year report SoftBank recorded a loss of 3.73 billion dollars, of which 2.85 billion in non-refundable advances.
A deposit can buy a date. The ability to choose later what to have produced can cost extra.
The distinction becomes very concrete in SoftBank’s half-year report for the period ended September 30, 2026. The group recorded charges of 3.73 billion dollars after a revision of purchases of power generation equipment intended for US data centers. Of these, 2.85 billion relate to non-refundable advances; 880 million to contract modifications and settlements. The report also indicates possible refunds of up to 2.1 billion over ten years, conditional on future purchases. SoftBank · Half-year report, p. 5
That possible recovery is not the same as money already back in the bank. It depends on later transactions and on conditions. In the accounts, precisely because the amount could not be determined, the right was not recorded as an asset. It is an important detail when you try to tell the cost of change with a single number.
We do not have enough elements to judge every one of SoftBank’s industrial choices from the outside. We do have a case in which the price of revision became visible. Technology can change quickly; commitments made to suppliers keep following the contracts signed.
The free week at the factory
Think of a packaging line. It is an example, on a scale closer to many companies. The supplier has a production slot available. The customer still has two configurations on the table and a technical check to complete. Paying the deposit makes it possible to keep the date, but it can also commit you to a machine that has not yet been fully chosen.
The first calculation is easy to recognize: price of the plant, installation, start-up, expected costs. The second requires taking seriously what we do not yet know. How much does it cost to change configuration? Does the deposit transfer? Until when can the quantity be reduced? What happens if the electrical check changes the design?
These questions do not express distrust of the investment. They make it possible to compare offers that otherwise look equivalent. A cheaper machine with rigid terms and a more expensive one that keeps a choice open have different profiles. The difference can be worth more than the discount negotiated on the initial price.
The investment committee also has to see the cost of the decisions it is bringing forward. If a piece of data will become available in a month, the contract can perhaps keep that part open until then. Or the supplier can ask a price for the flexibility. At least the choice becomes explicit.
Waiting can be the most expensive mistake
In markets with scarce capacity, however, waiting can compromise the project. If everyone waits for certainty, whoever books first gets supply and dates; whoever arrives later can lose customers, production and opportunities. A flexible clause is not much use if the supplier picks another buyer willing to commit right away.
Anyone who has lost a supply slot knows this risk. Turning every investment into a long wait does not resolve the uncertainty. The decision is about how much to pay to start and how much to keep freedom. In some conditions taking on a rigid commitment is the right choice, provided the expected advantage also justifies the risk of having to change it.
You can, however, separate decisions that are presented as a single package. Reserving production capacity, fixing a configuration and confirming the whole quantity can have different deadlines, if the supplier is willing to negotiate them. It is not always. The discussion has to take place before payment, when both sides still have something to trade.
The cost of standing still also has to be quantified consistently. An order lost because the plant arrives late is different from growth that is only hypothesized. Confusing the two can make any deposit look worthwhile. The comparison improves when benefits have a stated degree of certainty, like the conditions that can change the project.
Money already spent
After the deposit, the risk changes again. Part of the organization may want to go ahead so as not to make the loss visible. But the unrecoverable money is already committed: the next decision has to assess costs and results from that moment on. Adding a second expense to justify the first can make the mistake more expensive.
At that point you also need someone authorized to stop the project. If everyone answers only for sticking to the original plan, new information arrives without finding a possible decision.
Conditional refunds require similar attention. If recovering them means buying more, the new purchase has to have a reason of its own. Otherwise the recovery becomes an incentive to maintain a relationship or a configuration that is no longer needed. It is a point to read with someone who knows both the contract and the industrial plan.
The SoftBank case makes enormous something that also exists in smaller investments: changing your mind has a price, and that price is often set before the change becomes necessary.
In the meeting that authorizes the deposit, then, two pages are needed. One describes the project the company wants to carry out. The other makes readable what can still change, by when and at what cost. The best time to write the second is before signing the first.
Read next Scenari · Ingegneria aumentata con Capgemini (podcast, in Italian) · The test comes before the code
Image: SoftBank Group