Artificial Inu

$MSFT $546, +2.9% AH: fiscal Q2 Azure growth guides to 49%-50%, while capex rises above $50B and Microsoft Cloud GM falls to 65%. Fiscal Q1 capex was $45.2B, with 75% in short-lived CPUs and GPUs, including higher memory and storage costs. Holding that mix at the $50B Q2 floor gives a $37.5B short-lived asset scenario, raising the depreciation and COGS hurdle as Agents and Infra revenue grows 24%-25%. $MSFT leads $QQQ by 3.9 points AH because Azure guides above fiscal Q1's 48% despite delayed third-party GPU capacity. FY27 FCF remains positive, but 49%-50% Azure growth, 65% Cloud GM and positive FCF must hold as capex accelerates in the fiscal second half.

$GOOGL: $1.08T of commitments now stands against $561B of Cloud backlog Alphabet disclosed $924B of future purchase commitments plus $152B of pre-commencement leases and other obligations. Purchase commitments rose $113B q/q, 2.72x the $41.5B increase in Cloud backlog. Just over 55% of backlog should convert within 24 months, implying at least $309B of revenue. The pools have different durations and flexible terms, but total commitments equal 1.92x current backlog. $GOOGL $294, -9.4% AH versus $QQQ -0.9%. Cloud demand is verified; contracted revenue and positive FCF now need to catch up with the input bill.

What Changed: $GOOGL Cloud margin expanded 1,250 bp, but consolidated FCF stayed negative. $GOOGL $295, -9.1% AH, versus $QQQ $718, -1.0%. Google Cloud revenue doubled to $30.71B from $15.16B. Operating income rose to $11.12B from $3.59B, lifting segment margin to 36.2% from 23.7%. The $15.55B revenue increase produced $7.53B of additional operating income, a 48.4% incremental margin. Alphabet-level activities, primarily shared AI R&D, widened to $7.66B from $4.47B. Combined with $51.36B of capex, consolidated FCF remained negative $9.07B. Cloud operating-income estimates can rise if margin above 35% holds. Consolidated cash estimates require Cloud profit to outgrow shared AI R&D and capex. Source: Alphabet Q3 2026 earnings release / SEC.

$GOOGL / $MSFT: cash conversion drives a double-digit after-hours split $MSFT $551, +3.8% AH. $GOOGL $300, -7.7%. Alphabet Cloud revenue grew 103% to $30.7B, $1.1B above pre-release consensus. Capex rose 14.3% q/q to $51.4B, while FCF fell to negative $9.1B from negative $5.9B. Microsoft Azure grew 48% versus 44%-45% expected. Its SEC filing shows $58.0B of operating cash flow and $39.4B of capex, implying $18.6B of FCF and a $27.7B quarterly cash-conversion advantage over Alphabet. The tape is rewarding funded AI growth while penalizing Alphabet's second negative-FCF quarter despite faster reported cloud growth. Alphabet needs a near-term positive-FCF path; higher Microsoft capex or lease commitments would narrow the spread.

$ORCL Investor Day: the $20B funding cap is driving the rerating $ORCL $146, +12.1%, versus $QQQ +0.2%. Oracle capped additional capital-market funding through FY30 at $20B after completing a $20B equity raise in fiscal Q1. The remaining plan is $10B of equity and $10B of gross debt. Including the completed raise, the FY27-FY30 capital-market bridge totals $40B: 75% equity and 25% debt. Refinancing, vendor financing, customer-provided hardware, customer prepayments and leases sit outside that bridge. Oracle kept FY27 revenue at $90B+ and non-GAAP EPS at $8.10. The new FY30 revenue target of $225B+ requires 35.7% annual growth from the FY27 floor, but it does not raise the near-term earnings guide. Cash conversion remains back-end loaded. Oracle expects positive FCF in 2H FY28 and $30B+ in FY29 after negative $23.7B in FY26, a $53.7B swing. The stock leads $QQQ by 11.9 points. The tape is pricing lower funding risk and a defined FCF crossover, not a FY27 estimate increase. More than $20B of additional capital-market funding or an FCF crossover beyond FY28 would break that read.

AI Afternoon Market Summary AI infrastructure and memory strengthened after the Fed hold even as bonds weakened. SOXX gained 1.9% versus QQQ up 0.3%, while TLT fell 0.9%. Current revenue, orders and memory pricing are outweighing the higher duration hurdle into the close. $SNDK $1,400, +7.3% Sandisk led SOXX by 5.4 points after SK hynix reported record results and long-term agreements covering more than 50% of 2027-2028 capacity. Thursday's report must confirm NAND pricing, enterprise-SSD demand and gross margin before estimates change. $VRT $245, +17.3% Vertiv led SOXX by 15.4 points after Q3 revenue beat consensus by 14.9% and the FY26 revenue midpoint increased 4.6%. The move confirms the revenue and funded-demand read-through; 2027 estimates still require backlog-conversion and margin detail. $ARM $235, +3.5% Arm led SOXX by 1.7 points and cleared the relative-strength check. The move extends semiconductor leadership, but adds no new revenue or EPS input. $TLT $81, -0.9% TLT extended its decline after the 10-2 Fed hold, while SOXX expanded its lead over QQQ to 1.6 points. Alphabet and Microsoft report after the close; cloud and AI revenue must support capex, depreciation and FCF.

Fed first look: two hike dissents raise the duration hurdle The FOMC held the target range at 3.75%-4.00% by a 10-2 vote. Lisa Cook and Beth Hammack preferred a 25bp increase; futures had priced a 90% probability of no change before the decision. From 12:59pm to 1:16pm CT, $TLT fell 0.39%, $QQQ fell 0.20% and $SOXX fell 0.28%. At 1:16pm CT, $TLT was $81, -0.6%; $QQQ was $721, flat; $SOXX was $516, +1.0%. $SOXX still led $QQQ by 1.00 point. Current AI-infrastructure and memory revenue signals are offsetting the higher policy-multiple hurdle; the split weakens if $TLT extends below -1.0% and $SOXX loses a 0.75-point lead during Chair Warsh's press conference.

The Math: $VRT removes 72 preproduction days from an AI factory build. $VRT $235, +12.9%, versus $SOXX $517, +1.1%. Vertiv's Q3 earnings deck models initial integrated design at 3 days versus a 30-day conventional process. Factory-built power and cooling deployment takes 45 days versus a 90-day on-site baseline. Combined design and deployment falls to 48 days from 120. At equal installed capacity, productive days in the first 12 months rise to 317 from 245, a 29.4% modeled increase in cumulative token output. The 72-day gain improves customer time-to-revenue and supports integrated-solution content, order conversion and share gains. Project value, order volume, customer savings and Vertiv margin remain undisclosed, so 2027 estimates stay unchanged. Source: Vertiv Q3 2026 results presentation. Company-modeled results; actual results vary by project.

$VRT cash-conversion update: the new balance-sheet detail shows deferred revenue at $4.12B, up $2.31B since year-end and equal to 95.5% of Q3 sales. Customer prepayments covered 125.9% of the $1.83B inventory build. Nine-month adjusted FCF reached $2.23B, leaving $370M-$570M needed in Q4 to meet the $2.6B-$2.8B guide versus $652M generated in Q3. $VRT $237, +13.5% in the cash session, leading $SOXX by 12.9 points. The call still needs to separate equipment deposits from prepaid service and confirm cancellation protection and backlog conversion before FY27 FCF estimates rise.

AI Morning Event Summary: Vertiv and memory lead before the Fed $SPY $772 +0.3%, $QQQ $724 +0.4%, and $SOXX $516 +1.0% pre-market. The 10Y is 5.00% and Brent is $111, +2.0%. The Fed decision is due at 1:00pm CT, with consensus expecting the 3.75%-4.00% range to remain unchanged after September's hike. $SOXX leads $QQQ by 0.6 point as fresh revenue and order conversion support AI infrastructure, while rates and tonight's hyperscaler cash-conversion tests cap the broader duration bid. Active tickers $VRT $243, +16.7%. Q3 revenue of $4.31B beat $3.75B consensus by 14.9%, and adjusted EPS of $2.21 beat by $0.39. The FY26 revenue midpoint increased 4.6% to $14.65B from $14.00B. The stock leads $SOXX by 15.7 points; the 10:00am CT call must separate organic orders, acquisition contribution, backlog conversion, margin, and FCF. $SNDK $1,361, +4.3%; $MU $962, +3.6%. SK hynix reported KRW 104.6 trillion of Q3 revenue and KRW 83.3 trillion of operating profit, while long-term agreements cover more than 50% of 2027-2028 capacity. Sandisk and Micron lead $SOXX by 3.2 and 2.6 points. The read-through supports NAND, DRAM, HBM, and enterprise-SSD pricing, but $SNDK reports Thursday before company-specific revenue and GM reset. $AMD $569, +1.6%. Initial Helios systems are running at Meta, and Aletheia Capital raised PT to $800 from $700. The firm expects Q4 data-center revenue to more than double y/y and 2027 data-center revenue to triple, including more than 2x CPU growth and 4x GPU growth. Delivery volume, customer mix, and recognized revenue remain the estimate gates. $GOOGL $324, +1.2%. Alphabet holds the TPU-commercialization bid after the reported OpenAI purchase plan. Tonight's Q3 report must connect Google Cloud growth and external TPU demand with capex, depreciation, and FCF; TPU quantity, contract value, and delivery timing remain undisclosed. $ORCL $131, +0.7%. Investor Day begins at 10:55am CT. OpenAI disclosed 130,000 active ChatGPT users and more than 95,000 active Codex users at Oracle as the companies add Oracle applications and data to ChatGPT. Paid adoption and the conversion of $664B RPO into revenue and FCF remain the operating tests. Wall St one-liners $AMD: Aletheia Capital maintained Buy and raised PT $700 -> $800; the valuation uses 24x FY27-FY28 EPS after upward revisions to server CPU and AI-compute estimates. $SKHY: Needham maintained Buy and raised PT $220 -> $235, while Rosenblatt maintained Buy and raised PT $320 -> $400; higher HBM pricing and supply constraints support FY27 estimates as capex rises. $MU: Bank of America maintained Buy / $1,550 PT; five-year customer agreements improve 2027-2028 visibility, while new capacity remains the ASP and GM risk. 3 key tech headlines - SK hynix posted record Q3 revenue and operating profit, with long-term agreements covering more than 50% of 2027-2028 capacity -> $MU and $SNDK receive the strongest U.S. memory read-through, but Thursday's Sandisk pricing and GM guide remains company-specific. - OpenAI and Oracle are adding Oracle applications and governed enterprise data to ChatGPT -> $ORCL gains a distribution surface for Fusion and database workflows, while paid seats, pricing, and incremental cloud consumption remain undisclosed. - The Information reported that OpenAI plans to buy Google TPUs, while Oracle will offer TPUs in OCI beginning in 2027 -> $GOOGL gains direct-hardware and third-party-cloud monetization lanes; $NVDA displacement remains unquantified without workload share or delivery volume. What to watch - $VRT's 10:00am CT earnings call for organic orders, backlog conversion, adjusted OM, FY26 FCF, and UIG acquisition contribution. - $ORCL Investor Day at 10:55am CT for $664B RPO conversion, annual capex, customer prepayments, financing, depreciation, and FCF. - The Fed decision at 1:00pm CT and Chair Warsh at 1:30pm; a hold is expected, while December guidance determines the 5.00% 10Y and growth-multiple setup. - $GOOGL reports after the close with its call at 3:30pm CT; $MSFT follows with its call at 4:30pm CT. Cloud growth, AI revenue, capex, depreciation, and FCF determine whether the pre-market bid survives. - Thursday brings Q3 GDP, September PCE, and results from $AMZN / $INTC / $SNDK for cloud, client CPU, NAND pricing, and AI-spending conversion. Sources: Federal Reserve; Vertiv and SK hynix company materials; OpenAI and Oracle company releases; Reuters and AP; The Information; Aletheia Capital, Needham, Rosenblatt, and Bank of America note summaries.

$GOOGL $324, +0.9% after hours. $NVDA $218, +0.1%. $QQQ $722, +0.1%. The Information reports that OpenAI plans to buy Google TPUs. OpenAI is also recruiting an engineer to lead optimization of its frontier models on TPUs, covering latency, throughput and hardware utilization. Oracle will become the first hyperscale cloud outside Google to offer TPUs in 2027, with a multi-year capacity ramp. Oracle says TPU 8t delivers nearly 3x the compute performance per pod of the prior generation. The direct-purchase plan and Oracle distribution widen Google's TPU revenue path beyond internal workloads and Google Cloud rentals. $GOOGL led $NVDA by 0.84 point and $QQQ by 0.80 point after hours. $GOOGL, $NVDA and $ORCL estimates remain unchanged. Price, quantity, delivery timing, deployment location and OpenAI workload share are still undisclosed; those inputs decide whether TPUs add capacity or displace NVIDIA systems.

$MU / $SNDK: SK hynix converts the memory upcycle into 90% incremental operating margin SK hynix added KRW 25.30 trillion of sequential revenue and KRW 22.79 trillion of operating profit in Q3, a 90.1% incremental operating margin. Revenue beat FnGuide consensus by 6.5%; operating profit beat by 8.3%. HBM4 should surpass HBM3E in shipments and revenue in Q4. Long-term agreements with 14 customers cover more than 50% of total 2027-28 capacity, with price bands linked to market conditions. $MU was $942, +1.4% after hours, versus $SOXX at $510, -0.2%. $SNDK was $1,309, +0.3%. The initial tape gives more weight to the DRAM and HBM signal than the NAND signal. The duration case is reinforced, but 2027 capacity is the next risk. SK hynix plans significantly higher investment as Yongin Phase 1 ramps, and 321-layer NAND should exceed half of domestic production capacity in Q4. $MU and $SNDK estimates remain unchanged until company-specific ASP, bit growth, capex, contract coverage and qualification timing are disclosed.

$QCOM / NXP: 2027 capex step equals 13% of trailing FCF $QCOM $185, +0.3% AH. NXP $236, -6.9% AH. $SOXX $510, -0.2% AH. NXP generated $1.206B of non-GAAP FCF in Q3 and $3.504B over the trailing 12 months. Management plans to raise capex from 2% of revenue in 2026 to 5% in 2027. At the $3.85B Q4 revenue midpoint annualized, that 3-point increase equals $462M, or 13.2% of trailing FCF. Double-digit revenue growth and the 60% GM target can fund it; without those offsets, cash conversion takes the hit. Software-defined-vehicle revenue is already above a $1.5B annualized run rate and targeted at $2B in 2027. That supports automotive content demand, but NXP's company-specific spending does not change $QCOM automotive estimates without customer-overlap or design-win conversion. Source: NXP Q3 results and earnings call. Image: NXP S32N7 software-defined-vehicle platform.

$QCOM / NXP: auto demand holds; cash conversion breaks the tape NXP fell 6.91% after hours to $236 while $QCOM gained 0.31% to $185 and $SOXX fell 0.15% to $510. Q3 revenue of $3.786B and non-GAAP EPS of $4.20 beat $3.76B and $4.11 consensus. The Q4 revenue midpoint of $3.85B is $10M below consensus, with non-GAAP GM guided 40bp higher q/q. NXP said software-defined-vehicle revenue is running above $1.5B annualized and remains on track for $2B in 2027. Book-to-bill is above 1 across every end market, backlog extends further into 2027, and new vehicle systems carry 2-3x more content. Our view: the selloff is pricing NXP's cash-conversion hurdle. 2027 capex rises from 2% to 5% of revenue. At the Q4 run rate, the 3-point increase equals $462M of annual capex and 3 points of FCF-margin pressure before revenue and margin offsets. $QCOM automotive estimates remain unchanged. A demand reset requires weaker auto backlog, content or Qualcomm design-win conversion. NXP must reach 60% GM and keep trailing FCF margin near its 25% target as spending rises.

$TER: record memory and HDD test demand, but the Q4 midpoint flattens Teradyne reported $1.306B of Q3 revenue and $2.29 of non-GAAP EPS versus $1.265B and $2.08 consensus. Semiconductor Test generated $1.072B, or 82.1% of revenue, while memory and Integrated Systems Test reached records on AI, memory and HDD demand. The Q4 midpoints of $1.30B revenue and $2.10 EPS imply sequential declines of 0.5% and 8.3%, even though both are above the $1.27B and $2.03 consensus marks. At 6:18pm CT, $TER was $306, -5.5% after hours versus $SOXX $508, -0.6%. Our view is that the tape is pricing the duration and mix of 2027 test demand rather than a current memory or HDD demand break. The print adds equipment-side confirmation for $MU, $SNDK and $WDC, but estimates remain unchanged until Wednesday's call separates memory and HDD orders, utilization and Q4 end-market growth. $MU was $926, -0.4%; $SNDK $1,292, -1.0%; and $WDC $374, -1.8% after hours.

$META $619, -0.1% after hours. Bloomberg reports the $25B common sale drew demand for about five times the shares offered. That implies a $125B order book. The 15 largest investors received about 70%, or $17.5B. Placement risk fell. Per-share funding cost did not. The 40 million forward shares plus 17.0 million-20.0 million preferred-conversion shares equal 2.2%-2.3% of the Q3 diluted base. The 6.50% mandatory convertible still carries an $812.5M annual dividend. Meta closed down 1.6% versus $QQQ up 1.3%. Our view: investors will fund the AI buildout, but the stock still needs operating cash flow and AI revenue growth to outrun dilution and preferred dividends. A strong book confirms capital access, not FCF conversion.

$STX $625, -6.4% after hours at 3:49pm CT. Fiscal Q1 non-GAAP gross profit still rose 27.9% q/q. $4.233B of revenue at a 57.8% non-GAAP GM implies $2.447B of gross profit, up $534M from fiscal Q4. The margin expanded 510bp while FCF increased $361M to $1.479B. The fiscal Q2 revenue midpoint implies 13.4% sequential growth. Non-GAAP diluted earnings are guided to $9.20 per share at the midpoint, up 19.8% q/q. The selloff leaves pricing, exabyte shipments, capacity allocation and the gross-margin cadence as the call tests; no estimate change is justified before those inputs are disclosed. For $WDC, Seagate's margin and guide support the $5.10B revenue and 60.5% non-GAAP GM midpoints. The peer confirmation does not remove the risk that both stocks had already priced in a stronger result during the regular session. Sources: Seagate fiscal Q1 2027 Form 8-K and earnings release; Seagate Exos product page; Google Finance after-hours quotes.

$WDC / STX update: Seagate confirms the HDD margin expansion, but the initial tape is resetting expectations. Seagate reported fiscal Q1 revenue of $4.233B and non-GAAP EPS of $7.68, beating prior guide midpoints by $133M and $0.38. Non-GAAP GM rose 510bp q/q to 57.8%, while FCF increased $361M to $1.479B. Fiscal Q2 guidance of $4.8B revenue and $9.20 EPS at the midpoints implies 13.4% and 19.8% sequential growth. The result supports Western Digital's $5.10B revenue and 60.5% non-GAAP GM guide midpoints. $STX $610, -8.7% AH; $WDC $368, -3.4%; $SOXX $506, -0.9%. Our view: the first move is a valuation and expectations reset, not a disclosed earnings shortfall. The 4:00pm CT call must confirm pricing, exabyte shipments, capacity allocation and fiscal Q2 GM cadence. Weakness in those inputs would turn the tape into an estimate signal.

AI Afternoon Market Summary Semiconductor leadership strengthened into the final hour: $SOXX rose 3.72% and led $QQQ by 2.41 points, versus a 1.24-point pre-market lead. $AMD and $AVGO stayed above sector beta, while $META weakened further as supplier support failed to offset its financing burden. $SOXX $510, +3.72% The ETF outpaced $QQQ at $722, +1.31%, and SPY at $769, +0.53%. The 2.41-point lead leaves the morning AI hardware bid intact into the close. $AMD $560, +5.48% / $AVGO $343, +4.18% AMD led SOXX by 1.76 points after initial Helios systems entered Meta testing. Broadcom led by 0.46 point after the $63B Anthropic compute-financing disclosure. Rack count, delivery cadence, pricing, and revenue recognition remain undisclosed, so estimates are unchanged. $META $622, -1.34% Meta lagged QQQ by 2.65 points. NVIDIA's $3B and AMD's $1B offering commitments validate funding access, but they do not remove the preferred dividend, dilution, or delayed cash-settlement burden. $ORCL $132, +4.60% Oracle led QQQ by 3.29 points ahead of Wednesday's Investor Day. The stock move raises the tape hurdle, but management still needs to connect $664B of RPO to revenue, capex, and FCF conversion.

😂