Alibaba Holds $128 as Qwen Powers Apple Intelligence in China and Cloud Runs 38% Into the 17th Aug Print

Alibaba's US-listed shares trade at $128.93, up 1.28%, after a run that has taken the stock from $116.46 on July 23 to just under $129 in nine trading sessions — a 10.7% advance in under two weeks. The move accelerated on August 3, when the shares jumped more than 4% on the rollout of the company's largest AI model to date, and it has not paused since.

The distance travelled from the lows matters more than the daily change. In late July the stock was trading below its 21-, 50-, and 100-period simple moving averages, with the relative strength index in oversold territory at 29, the moving average convergence divergence below the zero line with a negative histogram, and on-balance volume still declining with no sign of accumulation. That was a chart in a confirmed downtrend after breaking below $104.78 and approaching the $90 support level on prior weakness.

Nine sessions later the same indicator sits above 80. That is one of the fastest momentum reversals available in a mega-cap name, and it happened without a single earnings datapoint.

The year-to-date picture explains why the bounce has room to run and why it remains fragile. Alibaba was down roughly 22% year to date heading into the August 3 surge, and carried an approximately 28% drawdown at the depths of the second-quarter decline. Against the all-time high of $319.32 set on October 27, 2020, the stock at $128.93 sits 59.6% below the peak. The all-time low of $57.20 was reached on September 29, 2015.

The analyst consensus is aggressively constructive and has been for the entire drawdown. Forty covering analysts carry a Strong Buy rating with a twelve-month average target of $189.72 — 47.08% above the current price. A separate compilation puts the average at $186.90, and an 82-analyst survey shows a six-month average of $188.90 with a range spanning $112.90 at the low to $260 at the high. The rating distribution runs 38 Buy, 1 Hold, and 1 Sell.

A stock trading 47% below consensus with 95% of analysts rating it a buy has either a credibility problem on the sell side or a genuine mispricing. The distinguishing test arrives on August 17.

RSI Above 80 Into A $128.04 Ceiling

The technical setup is the most stretched it has been in over a year, and it is running directly into resistance.

The immediate level is $128.04. Above that, the next resistance sits at $134.05 and then $140.67. Spot at $128.93 has just cleared the first of those three, which is constructive on a closing basis but leaves the pair of higher barriers between the current price and the round-number attention at $150.

The momentum reading is the problem. The relative strength index above 80 is deep into overbought territory on any conventional framework, and it got there in nine sessions from a reading of 29. A move of that velocity typically resolves in one of two ways: a sharp consolidation that relieves the condition while holding the breakout level, or a failed breakout that retraces the entire advance.

The distinguishing factor is what caused the move. Momentum built on flow rotation and headline enthusiasm tends to unwind fast. Momentum built on a fundamental repricing tends to consolidate and extend. This advance has elements of both. It captured enthusiastic trading that began July 8 on news that profits had become less volatile and instant-commerce losses had begun to decline, and it was amplified by a rotation of Northbound capital out of South Korean and Taiwanese chipmakers into Chinese internet names including Baidu, JD, and PDD Holdings.

That rotation component is the fragile part. Sector flows reverse without warning and without any change in company fundamentals, and a stock that rallied because capital left Taiwanese semiconductors can give it all back when capital returns.

Below spot, the levels that matter are $128.04 as the breakout retest, then the $116.46 area where the advance began, then $104.78 — the level whose break in the second quarter opened the approach toward $90. A rebound from the $71 to $80 support zone was flagged as possible at the depths of the decline, which frames how far this stock can fall when the downtrend reasserts.

The practical read is that $128.04 has to hold on any pullback for the breakout to remain valid. Losing it with the relative strength index unwinding from above 80 would confirm the move as a squeeze rather than a re-rating, and would put $116.46 back in play ahead of the August 17 print.

August 17 Is The Only Date That Matters

Alibaba reports its June-quarter results on August 17 before the US market opens, and the setup into that print is the single most important element of this forecast.

Consensus calls for revenue of RMB 268.86 billion, representing 8.6% year-over-year growth. That is an acceleration from the fiscal 2026 full-year growth rate of 2.74%, which took total revenue to RMB 1.02 trillion from RMB 996.35 billion. Full-year earnings fell 18.20% to RMB 105.90 billion.

The problem with the current setup is a mismatch between what the stock has been buying and what the report will contain. The 10.7% advance since July 23 has priced an AI infrastructure story — a new frontier model, an Apple integration, accelerating cloud growth, and a token-consumption ramp. The August 17 report will primarily disclose an e-commerce company's profit and loss statement, with instant-commerce subsidy costs and a capital expenditure line that is heading materially higher.

Both stories are true simultaneously. The question is which one the reported numbers make more salient.

The precedent from the March quarter is not encouraging. Adjusted earnings before interest, taxes, and amortization fell 84% year over year to approximately $740 million, and the company nearly broke even on an adjusted net profit basis. Net profit in the fourth quarter of fiscal 2026 dropped to RMB 86 million — a rounding error against a company generating over a trillion renminbi in annual revenue — driven by heavy investment and competition.

A repeat of that profit profile against a stock that has just run 10.7% into overbought territory produces an immediate and violent repricing. A demonstration that instant-commerce losses are narrowing while cloud accelerates does the opposite.

The market has already begun pricing the constructive outcome. The rally from July 8 was explicitly built on evidence that profits had become less volatile and that instant-commerce losses had begun to decline. That expectation is now in the price at $128.93, which means the August 17 report has to confirm it rather than merely fail to contradict it.

Twelve trading sessions separate this analysis from that release.

The March Quarter Showed Exactly What Instant Commerce Costs

The magnitude of the profit destruction from the on-demand delivery war deserves precise accounting because it is the variable that determines whether the AI story ever reaches the bottom line.

Adjusted earnings before interest, taxes, and amortization fell 84% year over year to roughly $740 million in the March quarter, a decline management attributed to expensive on-demand delivery. Segment losses tied to the instant retail push expanded by more than 500%. Net profit for the quarter came in at RMB 86 million, missing expectations significantly.

An 84% collapse in operating profit at a company with more than a trillion renminbi of annual revenue is not a margin problem — it is a business model under active assault. The subsidies required to compete in thirty-minute delivery against an incumbent with entrenched logistics density are structural rather than promotional, and they scale with volume rather than declining as share is gained.

The competitive damage is symmetric across the sector, which is the one piece of good news. The dominant local-commerce incumbent reported a full-year net loss of RMB 23.4 billion, roughly $3.25 billion, with its core local commerce segment swinging from profit to a RMB 6.9 billion operating loss — approximately $960 million — and its shares trading around 44% below the 52-week high. JD's diluted earnings per share were effectively halved, with the decline attributed explicitly to food delivery and new-business investment.

Everyone is losing money. That matters because a price war in which all participants are bleeding tends to end sooner than one where a well-capitalized incumbent can outlast a weaker challenger.

The regulatory layer accelerated that resolution. Chinese market regulators flagged the food-delivery price war as a priority enforcement case in late January 2026 and summoned seven platforms again on February 13, making direct consumer subsidies considerably harder to run. The spending migrated rather than stopped — one campaign in early February deployed 3 billion yuan, roughly $420 million, on a promotional giveaway that produced over 10 million free orders in nine hours, framed as AI agent adoption and thereby protected under a different policy framework.

Management has now set a timeline: positive unit economics in local services by 2027, with the focus shifting away from aggressive share acquisition. That is the pivot the stock is trading on.

RMB 80 Billion Bought Nine Points Of Share

The return on the instant-retail campaign is quantifiable and it is poor. Alibaba captured approximately nine percentage points of market share growth at a cost of roughly RMB 80 billion in subsidies, against an incumbent that retained around 60% share.

That works out to roughly RMB 8.9 billion — about $1.2 billion — per point of share in a category with negative unit economics at current pricing. Buying share at that cost only makes sense if the share is defensible once subsidies stop and if the lifetime value of the acquired customer exceeds the acquisition cost. Neither condition has been demonstrated.

The strategic framing has shifted accordingly. The company is now navigating what amounts to a sunk-cost position, having recognized that the subsidies were a necessary defensive maneuver rather than a sustainable growth engine. The purpose was protecting the core commerce franchise from a competitor moving up the value chain into general merchandise, not building a profitable delivery business.

That reframing is important for how the August 17 numbers should be read. If instant commerce is a defensive expense rather than a growth investment, then declining losses represent success rather than retreat, and the market should reward evidence of spending discipline over evidence of share gains.

The underlying pressure that made the defense necessary has not gone away. The core Taobao and Tmall franchise faces structural erosion in its dominance, with broader China consumer softness compounding the competitive dynamic. Revenue growth of 2.74% across fiscal 2026 on a base above RMB 996 billion is the arithmetic expression of a mature commerce business in a slow economy.

The consensus for the June quarter at RMB 268.86 billion and 8.6% growth implies a meaningful reacceleration from that full-year rate. Some of that comes from instant-commerce revenue consolidation — gross transaction volume that carries revenue but destroys margin — which means a revenue beat could arrive alongside a profit miss.

That is the specific trap in this print. Headline revenue growth accelerating while adjusted operating profit stays compressed would confirm that the company is buying revenue rather than earning it, and the stock at $128.93 with momentum above 80 is not priced for that outcome.

Cloud At 38% Is The Reason To Own It

The offsetting business is genuinely excellent and it is why the consensus target sits 47% above spot despite everything above.

Cloud Intelligence Group grew 38% in the March quarter, up from 36% in the December quarter, with artificial intelligence constituting 30% of external cloud sales. Growth has run above 30% for multiple consecutive quarters, and expectations point to acceleration toward 45%. AI-related revenue has posted triple-digit growth for eleven consecutive quarters.

An eleven-quarter streak of triple-digit AI revenue growth is a longer sustained run than almost any Western hyperscaler has produced, and it is happening at a company whose stock trades 59.6% below its all-time high. That gap between operational trajectory and share price is the entire bull case.

The commercialization infrastructure has been consolidated to support it. A dedicated business group was established in March 2026 bringing the model laboratory, the model-as-a-service line, and the Qwen business unit under unified command, with a stated target of exceeding $100 billion in annual cloud and AI commercialization revenue within five years. Token consumption for model services surged significantly across the three months disclosed on the fiscal 2026 earnings call.

The pricing strategy is aggressive. The cloud platform has rolled out tiered package discounts covering more than 150 models including the current flagship variants, explicitly designed to alleviate enterprise concerns about runaway inference costs. That is a share-capture strategy in a market where the constraint on adoption is cost predictability rather than model capability.

Aggressive pricing has a margin consequence, and it is the reason cloud growth of 38% has not yet translated into group profitability. Volume growth at declining unit prices requires scale to reach operating leverage, and the capital spending required to build that scale is covered below.

The comparison that frames the valuation question is the one between this cloud business and its US equivalents. AWS grew 37% to $42.23 billion in the June quarter at a 39.4% operating margin, and Amazon trades at a $2.99 trillion capitalization on the strength of it. Alibaba's cloud unit grows at a comparable rate on a smaller base with an unquantified margin, inside a company the market values at a fraction of that multiple.

Whether the discount reflects geopolitical risk, governance risk, or a genuine margin difference is the question the August 17 disclosure should help answer.

The Apple Deal Is Bigger Than The Revenue It Generates

The July 15 development is the most consequential strategic event for this company in years, and its financial contribution is almost beside the point.

China's internet content regulator registered Apple Intelligence for generative AI services, with Alibaba's Qwen large language model serving as the core engine integrated into China-market iPhones and other devices. The integration spans iOS, iPadOS, macOS, and visionOS for users in China, allowing access to Qwen's text and image capabilities from within Apple's own interface rather than through a separate application. The US-listed shares rose about 4% on the news.

The selection process is what makes this validating rather than merely lucrative. Alibaba reportedly beat competitors including Baidu and ByteDance on model capability, computing capacity, and business complementarity. Being chosen by the most demanding consumer hardware company in the world, over the two most obvious domestic alternatives, is a third-party assessment of model quality that no benchmark score can replicate.

The market context makes the win larger. Apple generated $20.5 billion in Greater China sales during its second quarter, up 28% from a year earlier, with China shipments up 24.4% and the company recently regaining the number two position in the Chinese smartphone market. Qwen is now the default AI layer on the fastest-growing premium handset installed base in China.

The strategic value is consumer distribution. Qwen has deep roots on the enterprise side through the cloud platform, but the company has faced severe challenges in the consumer market where competing assistants have captured mindshare. Embedding the model in Apple's operating systems solves that distribution problem without any customer acquisition spend.

Two caveats belong in any honest assessment. The arrangement is not exclusive — Baidu is also working with Apple on Apple Intelligence features for Chinese users, meaning the integration is a multi-vendor arrangement rather than a sole-source contract. And Apple unveiled a rebuilt Siri powered by a competing Western model at its developer conference this summer, arriving with the next operating system release this autumn, which sets a capability benchmark the China implementation will be measured against.

The revenue from an inference contract on a domestic handset install base, even at Apple's scale, is small relative to a company generating RMB 1.02 trillion. What it buys is credibility, and credibility is what closes a 59.6% discount to a prior peak.

Qwen's Efficiency Pitch Against Price Pressure

The competitive dynamic in Chinese AI has shifted from capability to cost, and Alibaba's positioning on that axis is the reason the August 3 model release moved the stock 4%.

The company rolled out its largest AI model to date specifically as a low-cost competitor pushed pricing lower across the market. That framing is important: the release was defensive as much as offensive, and the metric being marketed is efficiency rather than raw capability. The current generation runs with 14.3% fewer parameters and 8.7% lower active workload than its predecessor — specifications aimed directly at cost-sensitive enterprise buyers rather than at benchmark leaderboards.

Qwen is the most widely used AI application in China and Alibaba remains the country's largest AI company by any reasonable measure. That position was built through open-source distribution, which created a developer ecosystem that competitors have struggled to replicate, and the release of successive open-weight generations has strengthened it further.

The economics of an efficiency-led strategy cut both ways. Lower inference cost per token expands the addressable market and drives volume, which is what produced eleven consecutive quarters of triple-digit AI revenue growth. It also compresses revenue per unit of compute, which means the capital deployed to build capacity earns less per hour of utilization than it would in a market with pricing power.

The tiered discount structure across more than 150 models on the cloud platform accelerates that compression deliberately. Management has chosen volume and ecosystem lock-in over near-term margin, which is a defensible strategy in a market where the winner is determined by developer adoption rather than by quarterly profitability.

The risk is that price competition in Chinese AI has no natural floor. Multiple well-capitalized domestic players, several backed by strategic rather than financial imperatives, are competing for the same enterprise workloads with open-weight models that cannot be differentiated on licensing terms. That is a structurally deflationary market.

Whether 45% cloud growth arrives with expanding or contracting gross margins is the disclosure that matters most on August 17, and it is the one item consensus has the least visibility into.

Capex Is Going From RMB 126 Billion Toward RMB 380 Billion

The number that should temper enthusiasm about the AI story is the capital required to deliver it.

Fiscal 2026 capital expenditure ran RMB 126 billion — roughly $17.4 billion — up 46% year over year, with cloud infrastructure spending surging more than 200% year over year within that total. Future capital expenditure is expected to surge past RMB 380 billion as the company enters an intensified investment phase.

That is a tripling of the annual spend from a base that already grew 46%. RMB 380 billion is approximately $53 billion at current exchange rates, against a company whose full-year net income was RMB 105.90 billion and whose most recent quarterly net profit was RMB 86 million.

The comparison to the Western hyperscalers puts it in context and does not flatter the risk profile. Amazon guided 2026 capital spending to approximately $220 billion against trailing operating cash flow of $161.4 billion, and its trailing free cash flow has swung to an outflow of $7.6 billion as a result. Alibaba is proposing to spend roughly a quarter of that against a fraction of the cash generation, in a market with structurally lower pricing.

The strategic logic is that AI is now the priority for resource allocation, explicitly displacing instant retail as the first claim on capital. That reallocation is what the market has been asking for and is the reason the stock rallied. It does not change the arithmetic that the spending precedes the revenue by years.

The margin consequence has already been visible since the start of 2026, with high capital expenditure in AI combining with instant-commerce losses to compress profitability from both directions. Retreating from one front to fund the other reduces the compression without eliminating it.

The offsetting argument is the one every infrastructure buildout makes: data centers monetize over long lifespans, servers pay back within a few years, and revenue growth eventually outpaces capital intensity. That framework holds if utilization is high and pricing is stable. In a market where the flagship strategy is tiered discounting across 150 models to alleviate cost concerns, pricing stability is precisely the assumption in question.

An August 17 disclosure that raises the capital expenditure trajectory toward RMB 380 billion without a corresponding acceleration in cloud revenue would be the clearest available bear signal.

Chip Export Controls Are The Binding Constraint

Every element of the AI thesis depends on access to compute, and access to compute is subject to a policy the company does not control.

Further restrictions on advanced US chips would directly limit Alibaba's ability to expand the graphics processing unit infrastructure needed for its large-scale AI investment plan. That is not a hypothetical constraint — it is the reason a RMB 380 billion capital expenditure plan carries execution risk that a comparable Western plan does not.

The workarounds carry their own risk. Reports on July 31 indicated the company had routed 20,000 Nvidia chips to power a frontier model despite US export restrictions. Whether accurate or not, reporting of that character creates regulatory exposure in the jurisdiction where the shares are listed, and it arrives alongside a shareholder class action already filed against the company.

The strategic response has been domestic substitution and efficiency. Building models that run with 14.3% fewer parameters and 8.7% lower active workload is partly a cost strategy and partly a compute-scarcity strategy — a model that needs less silicon is a model that can be trained and served under export restrictions. That constraint has arguably made the engineering better.

The geopolitical overhang extends beyond chips. Chinese technology valuations carry a persistent discount reflecting delisting risk, variable interest entity structure risk, and policy risk that has no analogue in US-listed peers. That discount narrowed through 2026 as the domestic regulatory environment for technology companies improved, which is one of the reasons the consensus target sits at $189.72.

The counterweight is that global demand for artificial intelligence has proven relatively resistant to geopolitical blockades, and Chinese open-weight models have captured meaningful international developer adoption precisely because they are free of Western licensing terms.

For the forecast, the practical implication is a wider distribution of outcomes than the fundamentals alone justify. A stock with 38% cloud growth, an Apple integration, and a $189.72 consensus target trading at $128.93 is not mispriced by accident. The discount is compensation for a policy risk that cannot be modeled, and it will persist until it doesn't.

Key-Man Risk And The Qwen Leadership Departures

A specific execution risk emerged in March 2026 that has not been resolved and rarely appears in the bullish framing.

The technical lead of the Qwen model program resigned along with several researchers. Leadership disruption at the head of a frontier model effort raises genuine questions about roadmap continuity, and it arrived at precisely the moment the model ecosystem was gaining international traction.

Frontier AI research is unusually concentrated in small numbers of individuals. A dozen researchers departing a program of that scale can delay a model generation by quarters, and in a market where a competitor is pushing prices lower and Apple is measuring the China implementation against a Western benchmark, a delayed generation is expensive.

The counter-evidence is the August 3 release. Shipping the largest model to date, with measurable efficiency improvements over the prior generation, five months after the departures suggests the program has depth beyond any individual. The organizational consolidation in March — unifying the research lab, the model-as-a-service line, and the product unit under single command — appears designed precisely to reduce that dependency.

The broader governance question sits alongside it. Fiscal 2026 saw net profit fall 18.20% while revenue grew 2.74%, a shareholder class action has been filed, and the company is simultaneously running a defensive subsidy war, a capital expenditure program headed toward RMB 380 billion, and a model race against domestically and internationally funded competitors. That is three distinct capital allocation decisions being made at once by a management team that has changed composition repeatedly.

The stated commitment to positive unit economics in local services by 2027 is the accountability marker. If the August 17 report shows instant-commerce losses declining toward that target while cloud growth accelerates, management credibility improves and the multiple expands. If losses persist at the March-quarter run rate, the pivot narrative that drove this rally loses its foundation.

Execution, not strategy, is the variable. The strategy is sound and widely agreed. The execution record across fiscal 2026 was poor.

Valuation: A $189.72 Consensus Against A $112.90 Low

The dispersion in published targets is the honest summary of how little agreement exists on what this company is worth.

Forty analysts carry a Strong Buy consensus with a twelve-month average of $189.72, implying 47.08% upside from $128.93. A parallel compilation puts the average at $186.90. An 82-analyst survey shows a six-month average of $188.90 with a low of $112.90 and a high of $260 — a range spanning 130% of the current share price. One published framework carries a $230 target based on roughly 12 times fiscal 2028 earnings projections.

Systematic models run considerably lower and more chaotic. Monthly projections put August at $131.86, September at $135.36, October at $160.30, November at $211.47, and December at $194.01, with a year-end range of $190.13 to $197.89. Separate modelling expects 2026 outcomes anywhere from $102.64 to $165.72 on the constructive side, with a decline to $52.48 to $69.62 not ruled out. One projection places 2027 between $59.38 and $99.82, and another expects a rise toward $155.11.

A range from $52.48 to $260 for the same twelve-month horizon is not analysis. It is an acknowledgment that the outcome depends on binary events — the trajectory of the price war, the availability of compute, the regulatory posture in two jurisdictions — that cannot be forecast from financial statements.

The valuation anchor that carries the most weight is the earnings multiple. At 12 times fiscal 2028 projections the stock supports $230; at current pricing the market is applying a materially lower multiple to the same projections, which is the compression the bull case expects to unwind.

The counterweight is that fiscal 2026 delivered 2.74% revenue growth and an 18.20% earnings decline. Applying a growth multiple to a company that shrank earnings requires believing the shrinkage was investment rather than deterioration. That is exactly the question the June quarter answers.

The stock at $128.93 is priced for the AI story to work and for the subsidy war to end. Both are plausible. Neither is confirmed.

The Levels That Decide August

The forecast reduces to three levels and one date. Resistance sits at $134.05 and then $140.67, with $128.04 now functioning as the breakout retest immediately beneath spot. Clearing $134.05 on a closing basis opens $140.67 and puts the round-number attention at $150 in view, roughly 16% above current pricing and still 21% below the consensus target.

Support is $128.04 first — losing it invalidates the breakout with the relative strength index unwinding from above 80, which is the classic failed-squeeze configuration. Below that, $116.46 marks where this advance began, and $104.78 is the level whose break in the second quarter opened the approach toward $90. The $71 to $80 zone is the deep support identified during the worst of the decline.

The base case into August 17 is consolidation between $128.04 and $134.05. A stock that has run 10.7% in nine sessions with momentum above 80 does not typically extend without pausing, and there is no scheduled company catalyst between now and the earnings release. Systematic monthly projections centering on $131.86 for August sit squarely inside that band.

The bull path requires the June quarter to confirm three things: cloud growth accelerating from 38% toward the 45% expectation, instant-commerce segment losses narrowing materially from the March quarter's 500%-plus expansion, and capital expenditure guidance that scales with revenue rather than ahead of it. That combination takes $140.67 out and makes the $189.72 consensus target a twelve-month objective rather than a fantasy.

The bear path needs only a repeat of the March quarter. Adjusted operating profit down 84%, net profit near breakeven, and a capital expenditure line moving toward RMB 380 billion against 8.6% revenue growth would strip the AI premium out of a stock that has just added it in nine sessions. That resolves through $128.04 to $116.46 quickly.

Alibaba at $128.93 trades 59.6% below its 2020 peak, carries a 47% gap to consensus, runs the most-used AI application in China, powers Apple Intelligence on China-market iPhones, and reported RMB 86 million of quarterly net profit. The market has repriced it as an AI company. On August 17 it gets handed an e-commerce company's income statement.

That's TradingNEWS