Ellison scraps a 10b5-1 plan covering 4.7% of his Oracle stake after one day

Two prices sit in the same account of last week, and they cannot both be current. Sedaily values the 50 million shares in the terminated plan at $7.5 billion based on the closing price on the 11th ^[3], which implies about $150 a share ^[1]. The same story reports Oracle's close on the 10th at $341.39 ^[13]. At that price the 50 million shares come to $17.1 billion ^[2], and getting from there to $150 the next session takes a 56% fall ^[3].

The cross-checks favour the higher price. Oracle's market value is put at about $951.7 billion ^[14]. At $341.39 that works out to roughly 2.79 billion shares outstanding ^[4], and 38% of them, the stake Sedaily attributes to Ellison, is about 1.06 billion shares ^[4][5]. At $150 the whole company would be worth some $418 billion and his holding about $159 billion, against the $393 billion Bloomberg's index put on him at 10:10 that morning ^[8][16]. So what he cancelled was closer to $17 billion of stock than $7.5 billion.

The fraction is small either way. Fifty million shares is 4.7% of a 1.06 billion share holding ^[5], and selling every one of them would have left Ellison at about 36.2% of Oracle instead of 38% ^[6]. Against shares outstanding it is 1.8% ^[7], and the plan had about six weeks left to run when it ended ^[10].

Oracle said "no shares were sold under the plan, and Ellison has no other plan to sell Oracle stock" ^[2]. On motive, Sedaily points to the stock's weakness this year. That weakness reduced the appeal of selling ^[10]. The same 50 million shares were worth about $8.75 billion when the plan was adopted, before a 16% decline ^[8]. Oracle is down 23% for the year on concerns that heavy data centre investment is straining its finances ^[9]. The publication also cites foreign media reporting that Ellison views the current share price as undervalued ^[21].

None of that settles the spending. Sedaily reports that the withdrawal removes one source of uncertainty without erasing concern about rising AI investment ^[12]. The Wall Street Journal reported that OpenAI will buy $300 billion of computing power from Oracle over the next five years, an average of $60 billion a year ^[19][9], and that the deal requires about 4.5 gigawatts ^[19]. Oracle had already announced the $500 billion Stargate data centre project with Sam Altman and Masayoshi Son ^[20].

I think the reversal was a decision about price. An insider who thinks his stock is cheap has little reason to keep a schedule running that sells it, and 4.7% of a stake is a trim ^[5][21]. The counter-thesis is in Sedaily's own figures: if $150 is the live price, then this is the stock that fell 23% this year ^[9] and a 38% holder simply declined to sell into weakness ^[4]. That is a duller decision than declining to sell after a 41.36% session ^[13].