戈多 Godot
戈多 Godot
前 Manta Network 首席研究员，专注美股期权和基本面。
2Following
245followers
Feed
Livestream
Feed
AMD really won, wow
Q3 revenue 13.377 billion, year-on-year +45%;
Profit 3.923 billion, year-on-year +75%;
EPS 2.09, year-on-year +74%.
AI business revenue +97%, profit +168%, offsetting declines in PC client and gaming, sharply contrasting with Intel!!
Lip-Bu Tan @LipBuTan1 take a look at Lisa Su @LisaSu!!
——————
Of the new revenue, 99.3% comes from the AI data center business, with the biggest growth driver being CPUs.
The key point is that the AI business's incremental operating margin is 43%, meaning for every $100 of new revenue, $43 converts into new operating profit. This is very good data.
Moreover, this round of growth is supported by actual shipments.
In the earnings call, Lisa Su clearly stated that revenue growth mainly comes from more CPU shipments and sales of higher core count products, with price increases contributing relatively little.
Server CPU revenue is expected to double by 2027.
——————
The balance sheet is very healthy.
Cash and short-term investments 15.776 billion, interest-bearing debt 7.079 billion, net cash 8.697 billion.
Although 4.75 billion bonds were newly issued in August, debt maturities are mainly concentrated between 2029 and 2036.
Total debt to EBITDA ratio is only about 0.43x, interest coverage ratio 46x.
Debt repayment pressure is very low, fully capable of bearing the current interest burden.
——————
With such a good earnings report, why hasn't the stock price seen a larger revaluation?
According to a DCF model back-calculation, at a stock price of 510 and a relatively optimistic 10% discount rate, it implies AMD's full-year growth in 2027 of 83%.
This aligns with management's growth target of over 80%.
Based on the Q4 revenue guidance of $17 billion, AMD needs to grow about 14% each quarter in 2027.
This valuation also requires AMD to continue growing 35% in 2028, 25% in 2029, and eventually increase long-term free cash flow margin to about 30%.
In other words, if you buy AMD now, you are buying growth expectations beyond these figures for the coming years.
——————
The incremental growth mainly depends on GPU.
Lisa Su expects data center AI revenue to grow over 100% in 2027, with AI business revenue surpassing server CPU in Q4.
However, management has not disclosed specific GPU shipment volumes, average selling prices, or standalone profit margins.
So the growth expectations for this part actually leave room for stock price appreciation.
But this is hard to measure accurately, so I won’t casually post an AI result here.
On the day of AMD’s earnings report, regular trading rose 3.65%, and after-hours rose about 7%. This performance is already good but slightly below the roughly 9% two-way volatility priced in by the options market before the earnings.
In a market with stronger risk appetite, such earnings could easily lead to double-digit jumps or new highs.
——————
So the current market trading theme is not the AI revolution, but the AI capital expenditure return rate and sustainability.
In the credit market, AI corporate bonds have credit spreads significantly higher than ordinary investment-grade bonds.
Rising financing costs affect capacity expansion and order fulfillment.
At the macro level, high interest rates directly increase the discount cost of future cash flows. Widening credit spreads further raise the market’s required risk premium.
In the earlier DCF model, simply raising the discount rate from 10% to 11% is enough to reduce AMD’s theoretical valuation by about 14%.
Therefore, AMD’s current investment logic is quite clear.
On one hand, AMD needs to see if GPU actual shipments, operating profit, and cash flow can continue to exceed market expectations.
On the other hand, it needs to see if earnings upgrades can cover the valuation pressure caused by interest rates and risk premiums.
Of course, if the macro environment improves, $AMD’s upside elasticity will be unleashed.
戈多 Godot
I've actually been tracking Intel for a long time. This earnings report is really a bit disappointing. $INTC
Within 2 years, it can be removed from my watchlist.
The earnings section is made very complicated, with things like DCAI and CCPG. But in reality, Intel's business structure is like a seesaw.
On one side is revenue growth driven by AI development, and on the other side is the traditional fundamental business affected by AI development.
One reason for the sharp drop is the price increase in storage, which has put pressure on the PC business and impacted the core business.
Q4 revenue is expected to decline more than 20% quarter-over-quarter. Also, PC shipments in 2027 are still expected to decline by about 10%, and this pressure will continue into 2028. The management's tone is not about easing but about changes.
I'm honestly shocked.
If the AI revenue growth could cover the impacted fundamental business, that would be fine, but in reality, it does not.
The core DCAI business seems to be growing fast, but the CPU shipments, which account for 81%, only increased by 10%. The growth mainly comes from price increases.
Moreover, if the average price growth normalizes, 2027 will require stronger shipment growth to maintain revenue growth.
The foundry business is still bleeding money, and no other large external customers are in sight.
It's being squeezed tightly by TSMC and Samsung.
Crucially, management said in the earnings call that 2027 is a concentrated investment period, needing to simultaneously expand Intel 3 and 18A production, and prepare equipment for 14A production in 2028.
The company is very likely unable to achieve positive free cash flow in 2027.
Looking at it now, the CPU battle is decided. AMD and Nvidia are the winners, and as for ARM, we'll see in the upcoming earnings reports.
Amazon earnings report, no major issues.
No structural contradictions from AI eroding existing business for now.
Free cash flow negative $42.4 billion, capital expenditure expected to be $220 billion in 2026, and will further increase in 2027.
However, financial leverage is not high.
Cash and marketable securities $114.2 billion, interest-bearing debt $181.3 billion, net debt about $67 billion.
Total debt/EBITDA 0.99x, net debt/EBITDA only 0.36x.
Operating profit for the quarter is 16.6 times the interest expense, no debt repayment pressure.
For 2027-28, AWS is expected to maintain growth above 40%.
$AMZN
I've actually been tracking Intel for a long time. This earnings report is really a bit disappointing. $INTC
Within 2 years, it can be removed from my watchlist.
The earnings section is made very complicated, with things like DCAI and CCPG. But in reality, Intel's business structure is like a seesaw.
On one side is revenue growth driven by AI development, and on the other side is the traditional fundamental business affected by AI development.
One reason for the sharp drop is the price increase in storage, which has put pressure on the PC business and impacted the core business.
Q4 revenue is expected to decline more than 20% quarter-over-quarter. Also, PC shipments in 2027 are still expected to decline by about 10%, and this pressure will continue into 2028. The management's tone is not about easing but about changes.
I'm honestly shocked.
If the AI revenue growth could cover the impacted fundamental business, that would be fine, but in reality, it does not.
The core DCAI business seems to be growing fast, but the CPU shipments, which account for 81%, only increased by 10%. The growth mainly comes from price increases.
Moreover, if the average price growth normalizes, 2027 will require stronger shipment growth to maintain revenue growth.
The foundry business is still bleeding money, and no other large external customers are in sight.
It's being squeezed tightly by TSMC and Samsung.
Crucially, management said in the earnings call that 2027 is a concentrated investment period, needing to simultaneously expand Intel 3 and 18A production, and prepare equipment for 14A production in 2028.
The company is very likely unable to achieve positive free cash flow in 2027.
Looking at it now, the CPU battle is decided. AMD and Nvidia are the winners, and as for ARM, we'll see in the upcoming earnings reports.
The parent company of the New York Stock Exchange, ICE, is collaborating with Trade[XYZ] to introduce Hyperliquid perpetual contracts.
Over 99% of ICE's current clients are institutional, with limited reach to retail users.
Trade[XYZ] has development experience with @HyperliquidX.
OKX provides an entry point for retail users, reducing customer acquisition costs.
In June, ICE and OKX established a joint venture, OKXICE, each holding 50%, planning to offer ICE's financial products to their own clients.
The Trade[XYZ] perpetual contracts are also expected to be distributed through this joint venture platform.
For #OKX, the most direct commercial benefits come from participating in the brokerage business, trading service revenues, and future profit sharing of the joint venture.
Previously, ICE also proposed exploring with OKX the licensing of crypto asset price data, US regulated futures, NYSE tokenized stocks, and institutional-grade digital asset infrastructure cooperation. Future revenue sources may extend to data services, institutional trading, and technology services.
Star_OKX
An exciting step toward bringing perpetual futures into the regulated U.S. market.
Combining ICE’s institutional-grade market infrastructure, Trade[XYZ]’s onchain innovation, and OKX’s distribution and trading technology could help create a new generation of global, 24/7 financial markets.
Different technologies. Different strengths. One shared vision: making financial markets more open, efficient, and accessible.
Looking forward to building this together. 🚀
1⃣
Microsoft's core business has not yet been significantly impacted by AI.
Google's AI spending simultaneously supports the development of new AI businesses and maintains the competitiveness of search advertising.
However, the increased use of AI Mode is related to the slowdown in search ad growth.
For Microsoft's Office, Teams, and Microsoft 365, AI appears as a new value-added service.
Microsoft 365 commercial cloud revenue growth comes from both an increase in paid accounts and higher revenue per user.
Azure provides computing power service revenue, while Copilot and Agent leverage existing enterprise customer relationships to expand software revenue.
————————
2⃣
Microsoft has a more ample current cash flow buffer.
$58 billion operating cash flow, after expenses, still retains $18.6 billion in free cash flow.
Google's operating cash flow is $42.29 billion, expenses $51.36 billion, resulting in a negative free cash flow of $9.07 billion.
——————
3⃣
Both companies have large expenditures, but Microsoft can still cover capital expenditures with operating cash flow this quarter.
At the same time, an operating profit margin close to 50% provides a substantial profit buffer for future depreciation and operating cost increases.
戈多 Godot
1⃣
Google's total revenue, operating profit, and EPS for Q3 fell short of expectations.
Capital expenditure was 51.355 billion, operating cash flow was 42.29 billion, and free cash flow was negative 9.065 billion, marking two consecutive quarters of negative free cash flow.
Future purchase commitments stand at 924 billion.
————
2⃣
Search advertising is the fundamental business. Although AI has brought more queries, with year-over-year growth reaching a five-year high.
However, search revenue growth slowed from 17% last quarter to 12%.
Besides currency exchange rates and early World Cup spending, traditional search users have strong purchase intent and conversion motivation, so advertisers are willing to pay higher prices.
In AI Mode, users may ask consecutive questions or even have AI directly complete comparison, planning, and purchasing tasks.
It is currently uncertain whether the additional queries can generate the same level of advertising revenue.
————————
3⃣
Cloud growth is fast, and new profits are considerable, but the growth has not yet translated into a corresponding increase in free cash flow.
As a result, Google is using cash flow accumulated from its traditional search business and new financing to support a large-scale AI infrastructure expansion.
Investment is temporarily lagging behind output.
Hence, the stock plunged sharply after hours
1⃣
Google's total revenue, operating profit, and EPS for Q3 fell short of expectations.
Capital expenditure was 51.355 billion, operating cash flow was 42.29 billion, and free cash flow was negative 9.065 billion, marking two consecutive quarters of negative free cash flow.
Future purchase commitments stand at 924 billion.
————
2⃣
Search advertising is the fundamental business. Although AI has brought more queries, with year-over-year growth reaching a five-year high.
However, search revenue growth slowed from 17% last quarter to 12%.
Besides currency exchange rates and early World Cup spending, traditional search users have strong purchase intent and conversion motivation, so advertisers are willing to pay higher prices.
In AI Mode, users may ask consecutive questions or even have AI directly complete comparison, planning, and purchasing tasks.
It is currently uncertain whether the additional queries can generate the same level of advertising revenue.
————————
3⃣
Cloud growth is fast, and new profits are considerable, but the growth has not yet translated into a corresponding increase in free cash flow.
As a result, Google is using cash flow accumulated from its traditional search business and new financing to support a large-scale AI infrastructure expansion.
Investment is temporarily lagging behind output.
Hence, the stock plunged sharply after hours
Korean and Taiwanese semiconductors, capital flow over the last 7 trading days.
Samsung is the strongest, with foreign capital returning, net buying of 1.44 trillion KRW, and domestic institutions buying simultaneously.
SK Hynix is mainly supported by domestic institutions; foreign investors have cumulatively sold 810.5 billion KRW, while Korean institutions have bought 836.2 billion KRW, basically offsetting the selling pressure from foreign investors.
TSMC's capital structure is relatively weak.
Foreign investors have cumulatively sold about 64,900 lots; although investment trusts and proprietary traders continue to take over, the scale of buying is limited, and foreign capital has not yet shown a clear return.
Capital risk appetite has returned. Within HBM, there is a general optimistic view that Samsung will improve its market share in HBM4, taking share away from SK Hynix.
1/
HBM underlying DRAM technology paths diverge.
Samsung, SK Hynix, and ChangXin are advancing 4F².
Micron skips 4F², continuing to optimize 6F², and plans to directly shift to 3D DRAM in the future.
Samsung and Hynix are also developing 3D, with the divergence being whether to go through the 4F² generation.
2/
HBM upgrades have two main lines,
First, increasing storage density of the underlying DRAM.
Second, stacking and interconnects to improve capacity, bandwidth, and energy efficiency. This route dispute mainly focuses on the former.
4F² uses more complex manufacturing processes to achieve higher storage density.
Theoretically, cell area reduces by 33%, bit density increases by 50%.
The cost is that vertical transistor manufacturing is complex, capacitor spacing limits actual density gains, and yield poses challenges.
Micron uses mature 6F² to buy time, directly betting on 3D DRAM. The risk is that 3D DRAM commercialization is further away; if 4F² achieves low-cost mass production first, Micron may temporarily lose cost advantage.
3/
Background: Micron's CTO Scott DeBoer pointed out in a call that the company has researched 4F² technology for about 20 years and made multiple prototypes.
The final decision to abandon it mainly involves two issues.
First, manufacturing complexity and cost.
Second, Micron believes the key factor determining actual DRAM storage density is capacitor spacing.
Capacitors cannot shrink indefinitely, or else read signals and data retention capabilities will be affected.
The theoretical 33% area reduction of 4F² is difficult to directly translate into the same magnitude of total chip area reduction, and even harder to directly translate into manufacturing cost reduction.
4/
ChangXin has a special condition: relatively sufficient financing capability. IPO net fundraising was 66.3 billion yuan, and in September announced a 35 billion yuan R&D and testing investment plan, mainly using oversubscribed funds.
For ChangXin, continuous R&D on 4F² has the opportunity to achieve technological breakthroughs and market share, and its strategic value may be higher than short-term financial returns.
5/
At least judging from profit realization in 2027–2028, I support Micron's route more.
Micron expects its 1δ process launching in the second half of 2027 to retain the 6F² architecture while still achieving about 25% bit density improvement.
At the same time, there is already commercial revenue from HBM4, allowing continued investment in yield and packaging improvements.
Micron can continue to reduce costs without bearing the extra risks of 4F².
4F² is unlikely to immediately change the HBM market supply-demand landscape.
$MU
