The obvious reading of Hock Tan’s Monday line is that Broadcom told the market the AI trade is a casino, so the stock should have stayed down. It did not. Nasdaq’s official Tuesday close, 13 October 2026, was $357.31, up $2.58, or 0.73 percent, from Monday’s $354.73. The session opened at $349.595, under that Monday close, and still finished higher. The casino headlines miss the map. A path through 31 December puts the bear at $264.27, within $1 of a $265 bear already on FinanceFeeds, and a 9 December horizon puts the bear at $279.66, within $1 of a published $280 bear. Neither pair is reused. On the 10 December date one calendar projects, and labels as unconfirmed, the last 90 daily log-returns give a bull of $433.07, a base of $332.34 and a bear of $278.56. The base is below $357.31. Tuesday is a close under the 2 June high of $481.57, not a floor, and it is not the 29 September page that still carries a $715 bull.
Key facts
- Tuesday 13 October 2026 regular-session close $357.31, up $2.58, or 0.73 percent, from Monday’s $354.73. Open $349.595, high $359.25, low $348.1101, volume 17,266,690. Source: Nasdaq historical prices, downloaded 14 October 2026. The file’s highest close is $481.57 on 2 June. The intraday high is $495.00 on 3 June, matching the quote page’s 52-week range of $289.96 to $495.00.
- At 7:26 a.m. Eastern on Wednesday 14 October, Nasdaq still marked the session premarket. Last sale $355.49, down $1.82, or 0.51 percent. Source: the Nasdaq quote saved that morning. It is not a close and not a chart bar. Average volume on that summary was 22,892,538. Tuesday’s 17,266,690 shares were under it.
- Hock Tan, president and chief executive, in Singapore on Monday 12 October, as Bloomberg reported via Yahoo Finance: “A lot of investors have no clue what they are investing in.” They “set very extreme expectations only to find out that it’s almost like they are in a casino.” On OpenAI’s roughly $50 billion September run rate, he said the lab is “still losing money” because it is “spending a lot on training.”
- Elon Musk, on X at 16:30 GMT on 12 October, replying to @StockMKTNewz: “No crying in the casino.”
- The 2 September earnings exhibit, SEC Exhibit 99.1, is unchanged by Monday. Q3 revenue was $29,591 million. AI semiconductor revenue was $16.7 billion, up 221 percent year over year. Q4 guidance, for the quarter ending 1 November 2026, is about $34.8 billion of revenue and $21.7 billion of AI semiconductor revenue. The $0.65 dividend was payable 30 September to holders of record on 21 September.
- The Form 10-Q for the quarter ended 2 August puts remaining performance obligations at about $179.2 billion. About 25 percent, or $44.8 billion, is expected over the next 12 months. Purchase commitments total $126,821 million. A customer-lease backstop has a maximum potential liability of about $29 billion. Nothing had been paid. Shares outstanding were 4,773,629,865 as of 28 August 2026.
- Nasdaq’s summary, saved before Wednesday’s open, listed market capitalization at $1,697,884,670,383 and a one-year target of $530.00. That street figure is not a case below. The annualized dividend was $2.60, a 0.73 percent yield, ex-dividend 21 September, already past. The yield matches Tuesday’s percent gain and does not explain it.
- Rich Privorotsky, head of European One Delta trading at Goldman Sachs, told MarketWatch, carried by Morningstar on 13 October, that the gambler line was “taken wildly out of context” and that Broadcom and Intel had reaffirmed “exceptionally strong demand for another 2-3 years, potentially through 2030.”
What just happened, and why the obvious reading is wrong
Monday’s tape did fall. Friday 9 October closed at $361.54. Monday closed at $354.73, down $6.81, or 1.88 percent, on 16,336,770 shares, with the close on the $354.22 low. A reader who stopped there could treat the casino line as a demand cut. The next official close does not.
Tuesday opened at $349.595, about $5.14 under Monday, and traded to $348.1101. That was the weak print. The session then closed at $357.31, above Monday, on 17,266,690 shares. The high, $359.25, did not retake Friday. From the 6 October close of $375.81, the level in an earlier FinanceFeeds note, Tuesday is down $18.50, or 4.92 percent. The bounce repaired Monday. It did not repair the week.
MarketWatch, filed at 4:29 a.m. Eastern on 13 October, said the stock had fallen another 1 percent on Tuesday. That line was written before the cash close. The official close was up $2.58. The morning line is not the session.
The June high is the other mistake. The highest close in the file is $481.57 on 2 June. Tuesday is 25.80 percent under that close and 27.82 percent under the 3 June high of $495.00. It is 23.23 percent above the 30 March low of $289.96. A price under the high is not a base. The base calculated here is $332.34.
Who is exposed
Holders are first. Nasdaq listed market capitalization at $1,697,884,670,383 against the $357.31 previous close. The 10-Q share count, 4,773,629,865 as of 28 August, should not be multiplied by Tuesday’s price into a second market cap. A $2.58 move does not reprice the order book.
Customers are the second exposure. Semiconductor solutions were $20,839 million, 70 percent of the quarter, and infrastructure software was $8,752 million, 30 percent. AI semiconductor revenue of $16.7 billion was 80.1 percent of the semiconductor segment and 56.4 percent of the $29,591 million total. The $21.7 billion Q4 guide would be 62.4 percent of the $34.8 billion revenue guide. The 10-Q describes “unprecedented demand” from customers including “the leading frontier AI labs,” and says meeting it “requires our customers to access significant capital.” That capital line is already in the filing. It is not a new cut to $21.7 billion.
The desks long the name are long a tape whose 90-session drift is negative, from the 4 June close of $418.91 to $357.31. Nasdaq’s $530.00 one-year target is 48.3 percent above Tuesday. The 29 September FinanceFeeds page used a $715 bull and a $216 bear. Those levels are not the cases here.
What the tape and the filings show
Change is versus the prior close in the Nasdaq file. Wednesday 14 October had not produced a cash close when the file was saved. The 7:26 a.m. Eastern premarket sale is not a row.
| Session | Close | Change | Volume | Range |
|---|---|---|---|---|
| Tue 13 Oct 2026 | $357.31 | +$2.58, +0.73% | 17,266,690 | $348.1101 to $359.25 |
| Mon 12 Oct 2026 | $354.73 | -$6.81, -1.88% | 16,336,770 | $354.22 to $360.6899 |
| Fri 9 Oct 2026 | $361.54 | +$1.40, +0.39% | 16,843,810 | $361.10 to $366.63 |
| Thu 8 Oct 2026 | $360.14 | -$16.37, -4.35% | 27,145,240 | $357.4101 to $373.35 |
| Tue 6 Oct 2026 | $375.81 | +$13.30, +3.67% | 30,834,300 | $364.01 to $380.84 |
| Thu 4 Jun 2026 | $418.91 | start of the return window | 80,979,450 | $403.0101 to $426.48 |
| Wed 3 Jun 2026 | $479.23 | intraday high $495.00 | 50,126,450 | $472.64 to $495.00 |
| Tue 2 Jun 2026 | $481.57 | highest close in the file | 38,306,650 | $470.459 to $488.82 |
| Mon 30 Mar 2026 | $293.41 | intraday low $289.96 | 28,054,450 | $289.96 to $304.77 |
Thursday 8 October, not Monday, was the heavy down day: $360.14, down $16.37, or 4.35 percent, on 27,145,240 shares. Tuesday’s volume sat under the 22,892,538 average. The $21.7 billion AI guide, the $179.2 billion of remaining obligations, and the $126,821 million of purchase commitments were on file before Monday and were not revised.
What Tan said, and what the company has not said
“A lot of investors have no clue what they are investing in,” Tan said in Singapore, in Bloomberg’s wording. They “set very extreme expectations only to find out that it’s almost like they are in a casino.” On OpenAI he said: “For now it’s still losing money because they’re spending a lot on training,” and the run rate “doesn’t tell anything at this point except that there is a lot of potential revenue growth out of providing frontier generative AI models.” The $50 billion figure is a revenue observation, not a profit. Protos reported the arcade line as well: “Your parents give you $10 of tokens, you finish it in 10 minutes and you haven’t learned anything.”
That is a complaint about buyers of the theme, and about labs that spend the revenue on training. It is not a sentence that cuts the Q4 AI guide. On 2 September, in Exhibit 99.1, Tan said demand for custom AI accelerators and networking “continues to be very strong,” cited the $16.7 billion quarter, up 221 percent year over year and 54 percent quarter over quarter, and said: “In Q4 the momentum continues, and we expect AI semiconductor revenue to accelerate to $21.7 billion, up 236% year-over-year.” No later filing fetched here replaces that.
Amie Thuener, chief financial officer, said in that exhibit that non-GAAP operating income rose 92 percent and that Q4 revenue was forecast up 93 percent to $34.8 billion, with the non-GAAP margin held at 66 percent. Sixty-six percent of $34.8 billion is $22.968 billion. Free cash flow was $13,665 million, 46 percent of revenue. The release puts headquarters in Palo Alto. The photograph with this article is the San Jose campus, taken 1 June 2019.
No transcript of the panel is in the documents used here. MarketWatch said it asked Broadcom and DBS for a recording and did not have one. Intel chief executive Lip-Bu Tan was on the panel. Wolf Richter, of Wolf Street, took the other side in the same MarketWatch story: the lab model, the corporate return, and the financing “remain uncertain,” and financing is “the biggest constraint.” Musk’s four words agree with the metaphor. They are not a guide. Reuters’ X post said Aoifinn Devitt of Moneta called the analogy too far. That is a paraphrase, not a transcript.
Capital is the constraint the filing already named
Purchase commitments in the 10-Q, unconditional and longer than a year, total $126,821 million. The rest of fiscal 2026 is $1,110 million. Fiscal 2027 is $52,674 million. Fiscal 2028 is $72,952 million. Fiscal 2029 is $85 million. Cancelable agreements and commitments of a year or less are excluded, so $126.8 billion is the part the company treats as binding. A panel line does not cancel it. A capital shortage can delay it.
In that quarter a financial partner took on AI-rack purchases and the related customer leases. On a default the liability is the gap between 85 percent of the amounts outstanding and the sale value of the racks. The maximum, once every rack is deployed, was about $29 billion, and nothing had been paid. The same note describes the AI XPV platform, a June 2026 tranche of $35 billion for more than 1 gigawatt, toward more than 20 gigawatts through 2028. Bloomberg also reported that Firmus Grid withdrew a planned $5.5 billion offering, and that DBS chief executive Tan Su Shan said the market is repricing AI-financing risk. None of that revises $21.7 billion.
Bull, base and bear
These are not bank targets, and they are not the $530.00 on the Nasdaq page. The last 90 daily log-returns run from the session after the 4 June close of $418.91 through Tuesday. The mean, mu, is -0.0017672483. The sample standard deviation, sigma, is 0.0275655662. The 3 June high of $495.00 sits just outside the window. The 90 returns still compound to the drop from $418.91 to $357.31, because they telescope.
Through 31 December, 55 sessions after Tuesday once Thanksgiving and Christmas are removed, the bear is $264.27, within $1 of the $265 bear on the August FinanceFeeds page. That bull was $560. Through 9 December, the date earnings.report prints, 40 sessions give a bear of $279.66, within $1 of the $280 bear on the 19 September page. Neither pair is reused. Next Earnings Date, saved 14 October, projects 10 December after the close and says the company has not confirmed it. The prior fourth-quarter report on that calendar was 11 December 2025. The release dates the quarter’s end as 1 November, a Sunday, and names no reporting day. N below is 41 sessions from 14 October through 10 December, weekdays only, with 26 November removed.
The square root of 41 is 6.4031. Sigma times that root is 0.1765 in log space, about 17.65 percent. One and a half standard deviations are 0.2648. The drift, N times mu, is -0.0725. Base is $357.31 times e to the power of that drift, $332.34, 6.99 percent under Tuesday. Bull adds the 1.5 standard-deviation term, $433.07, 21.20 percent above Tuesday and still 10.07 percent under the 2 June close of $481.57. Bear subtracts one standard deviation, $278.56, 22.04 percent under Tuesday. This is not financial advice.
| Case | Price | Versus $357.31 | What has to be true |
|---|---|---|---|
| Bull | $433.07 | +21.20% | The drift continues and the 10 December close finishes 1.5 standard deviations above it. The $21.7 billion guide holds. The path does not retake $481.57, the $530 target, or the $715 bull. |
| Base | $332.34 | -6.99% | Only the mean continues. Tuesday’s bounce is not a floor, and the guide is neither raised nor cut. $357.31 sits above this price, so it is not the base. |
| Bear | $278.56 | -22.04% | The drift holds and the close finishes one standard deviation under it, through a slipped guide or a wider backstop. This bear is not the published $265 or $280. |
$433.07 is still $48.50 under $481.57. The drift since 4 June points at $332.34, not at $357.31. The $715 page was a different calculation.
What happens next
Wednesday 14 October is the first check. At 7:26 a.m. Eastern the quote was still premarket, last sale $355.49, down $1.82. If the cash open holds under $357.31, it matches Tuesday’s open at $349.595, not a new base. Only the official close confirms or cancels the bounce. A close back through Monday’s $354.73 would say Tuesday was a one-session repair.
The quarter Broadcom has dated ends 1 November 2026. The guide is about $34.8 billion of revenue and $21.7 billion of AI semiconductor revenue. Nothing fetched from the company between 2 September and this morning revises either number. The Singapore remarks become a guide cut only if a filing or a transcript says so.
The projected report is 10 December 2026, after the close, and the calendar says it is unconfirmed. Another calendar prints 9 December. The test is whether Q4 AI semiconductor revenue is still $21.7 billion. A print at the guide would leave the negative drift as a tape fact. A cut would be the first primary sentence this story does not have. The year-end path was set aside because $264.27 sat too close to a published bear, and that path still does not retake $481.57.
Frequently asked questions
Is $357 a base under the June high?
No. Tuesday’s official close of $357.31 is 25.80 percent under the 2 June closing high of $481.57, and the base case through the projected 10 December report is $332.34, not $357. The bull case, $433.07, is still 10.07 percent under that June close. The September page with a $715 bull is a different map, and it is not reused.
What did Hock Tan say, and what did the stock do the next day?
On Monday 12 October in Singapore he said a lot of investors “have no clue what they are investing in” and that extreme expectations felt “almost like they are in a casino,” Bloomberg reported. Monday’s close was $354.73, down 1.88 percent. Tuesday opened at $349.595 and closed at $357.31, up $2.58. The weak print was the open. The official close was higher.
Did Broadcom cut its AI semiconductor guide?
Not in any document fetched for this piece. On 2 September Tan guided Q4 AI semiconductor revenue to $21.7 billion, after a Q3 result of $16.7 billion. The 10-Q, for the quarter ended 2 August, still describes unprecedented demand and says customers need capital. A panel remark about investors is not a revised guide. No Singapore transcript was in the sources used here.
What are the bull, base and bear prices?
Through the projected 10 December 2026 close, using the last 90 daily log-returns and 41 sessions, the bull case is $433.07, the base case is $332.34 and the bear case is $278.56, against Tuesday’s $357.31. Bull is 1.5 standard deviations above the drift. Base is the drift alone. Bear is one standard deviation below it. They are not the $530 Nasdaq target, and they are not financial advice.
Why not the year-end or 9 December levels?
Through 31 December the bear is $264.27, within $1 of a $265 bear already published with a $560 bull. Through 9 December, the date one calendar prints, the bear is $279.66, within $1 of a published $280 bear. Both pairs are unused. The 10 December projection is the horizon that stays clear of those pairs. Broadcom has not confirmed any of the three dates as its reporting day.
What is the capital figure that actually sits in the filing?
Purchase commitments total $126,821 million, mostly fiscal 2027 and 2028. The lease backstop’s maximum potential liability is about $29 billion, and nothing had been paid. June’s AI XPV tranche was $35 billion for more than 1 gigawatt. Remaining performance obligations were about $179.2 billion at 2 August, with about $44.8 billion expected inside 12 months. The casino line does not replace those figures.