Software & Cloud Sector Weekly (2026-09-28)

Report date: 2026-09-28 | X search window: 2026-09-21 to 2026-09-28 (industry-signals section extended to 2026-09-14) Price data as of: US equities 2026-09-25 close (US market not yet open on 09-28); HK equities 2026-09-28 close. All single-stock moves verified via authoritative quote snapshots — no price figures were taken from X post bodies. Sources: grok-x live X search (6 calls across broad / KOL / media / industry-event tracks), eastmoney quote snapshots; institutional research cited with source and date.

1. Key Events

1. Oracle issued a "force majeure" notice on its New Mexico data center project — Project Jupiter, built to serve OpenAI. The natural gas pipeline required to power the site had its permit denied by the state, pushing energization from August 2026 to February 2027. Per Bloomberg, ORCL fell as much as 5.4% pre-market on Sept 24. The single heaviest event in software & cloud this week. 𝕏¹ 𝕏²
2. Zscaler CRO Mike Rich stepped down for personal reasons, with global sales head Ross Tackett taking over effective Oct 1. ZS closed Friday at $193.05, down 10.06%, dragging the broader cybersecurity group with it. 𝕏
3. Microsoft unveiled a revamped Copilot app adding a coding tool and an always-on AI agent. MSFT closed Friday at $516.17, up 3.66% — the strongest large-cap software name of the week. 𝕏
4. Stifel upgraded Microsoft to Buy with a $575 price target, citing accelerating Azure growth, continued Copilot adoption and rising GitHub usage. 𝕏¹ 𝕏²
5. Microsoft began a new round of layoffs affecting roughly 500 employees, concentrated in Xbox gaming but also touching Cloud and AI units (post cites Business Insider). 𝕏
6. Oracle expanded its planned job cuts by approximately $700 million (Sept 14). 𝕏
7. Sharp intra-sector divergence — with a reversal on Friday: Datadog closed at $268.13, up 4.36%, and Snowflake rose 0.58%, while CrowdStrike fell 2.90%, Cloudflare fell 2.73% and HubSpot fell 3.53%. 𝕏

2. Institutional & Media Coverage

Factual reporting

• @DeItaone (relaying Bloomberg): Two successive headlines confirmed Oracle sent a force majeure notice to a Blue Owl-affiliated New Mexico data center developer; shares fell 5.4% pre-market, then 5.1%. The causal chain is clean: power/pipeline permitting blocked → project delayed → Oracle asserts contractual relief. 𝕏¹ 𝕏²
• @unusual_whales: "Oracle, $ORCL, is expanding its planned job cuts by approximately $700 million" — confirmed Sept 14. 𝕏
• @DeItaone: An earlier Sept 14 headline read "ORACLE STARTS A NEW ROUND OF LAYOFFS: INSIDER" — insider-sourced and therefore rumor-grade, subsequently corroborated by the budget figure above. 𝕏
• @wallstengine: Microsoft began cutting roughly 500 roles on Tuesday, mainly in Xbox but also in Cloud and AI. 𝕏
• @wallstengine: Satya Nadella said Microsoft's two-decade presence in Quincy, Washington grew local tax revenue roughly 12x and employs over 1,000 people. In his words: "Ultimately, if you're contributing to the local economy and doing it in a way that benefits the people there, I think we'll have the permission." Placed in the same week Oracle sought relief over blocked permitting, both point at one new variable: local permitting and community acceptance are now a binding constraint on cloud capex. 𝕏
• @EngageProVideo (via CNBC): Confirmed the project name as Project Jupiter and said Blue Owl, the developer's parent, "slipped too." Verified: Blue Owl (OWL) closed Friday at $9.32, up 0.76% — inconsistent with "slipped"; the quote data governs. 𝕏

Views / ratings

• Stifel (via @wallstengine): Upgraded MSFT to Buy, $575 target, expecting mid-to-high-teens revenue growth, and argued open-weight models actually strengthen Microsoft's LLM-agnostic AI strategy across Azure and Copilot. 𝕏
• Sell-side on the Zscaler transition (via @Blackintus, @Morpheu5Watcher): Guggenheim, Rosenblatt and Mizuho all framed it as "continuity, not disruption"; Mizuho raised its target to $220 and RBC to $236. Unanimous sell-side reassurance against a single-day 10.06% decline is the week's clearest news-versus-price contradiction. 𝕏¹ 𝕏²
• UBS (via @OptionKing666): Noted Oracle "is actively seeking to delay payments if the facility is unable to come online in CY28," and cited a sector-level figure — public pushback stalling or cancelling roughly 45 data center builds worth $68B. 𝕏
• Morgan Stanley (via @gin_cryptodiary): Warned the move "opens credit risk across AI infrastructure debt, especially for speculative builds facing rising financing costs and community pushback." Second-hand relay; no primary research note verified. 𝕏

Note: No qualifying original software/cloud posts were retrieved this week from @Reuters, @CNBC, @WSJ, @FT, @business, @FactSet, @TheTranscript_, @Quartr_App or @eWhispers. The China-assets KOL and media whitelists (@HAOHONG_CFA, @yicaichina, @cnfinancewatch and others) returned zero results on Chinese software/cloud/SaaS topics. This section is kept short accordingly rather than padded.

3. KOL Bull & Bear Views

Oracle / AI infrastructure credit chain

Bearish (no bullish posts were found on this topic this week)

• @slamwontmiss (≈7 likes, ≈720 views): "Oracle just declared force majeure on its New Mexico data center. The problem? The natural gas pipeline needed to power the facility has been delayed after its permit was denied amid legal challenges. No pipeline means no gas for the generators. Oracle took out an $18 billion loan... debt was already trading at 88–90 cents on the dollar... Now Oracle is telling Blue Owl, 'We're not going to pay you,' while CDS on Oracle's debt is skyrocketing." Frames the event as an escalation from project delay to credit event, flagging forced institutional selling if Oracle loses investment grade. 𝕏
• @MasterOfNonees (≈1 like, ≈100 views): Lists FY27Q1 figures — "Capex: $28.5B / Operating cash flow: $23.1B / Free cash flow: negative $5.4B / Total debt: $125B... backlog is $664B. $300B of it is OpenAI" — and says the stock fell to $139, down ~30% year to date. These financials come from the X post and were not cross-checked against the 10-Q; cited as opinion material only. The $139 reference corresponds to around Sept 24; the verified Sept 25 close was $137.10. 𝕏
• @JF_Stanley (≈3 likes): A one-line framing — "first crack in the AI capex supercycle?" — capturing the bulls' core worry. 𝕏
• @stock_duty: "if the AI buildout's chief vendor is lawyering out of overruns, every neocloud and data-center financier gets repriced. Watch $CRWV sympathy." Verified: CoreWeave closed Friday at $87.59, down 2.82%, directionally consistent. 𝕏

SaaS structural divergence / AI agents vs. the business model

Bullish (infrastructure and security layer)

• @HexxRL (≈13 likes; second post ≈19 likes): "Market divergence in SaaS: Infra & Security ($CRWD, $NET, $SNOW) sit near 52w highs, while core application platforms ($ADBE, $INTU, $HUBS, $CRM) remain heavily compressed on seat-decay fears. LLMs provide reasoning, while SaaS provides state, RBAC, and governance. Autonomous agents don't replace SaaS, they use it as their execution tool layer." The thesis is "Agent Gravity" — agents need SaaS for state, permissions and governance, expanding rather than eroding gross margin. Friday's tape reversed the divergence: the CRWD (-2.90%) and NET (-2.73%) he calls near highs fell more than INTU (-0.48%) and ADBE (-1.45%), which he calls compressed. The framework got no price confirmation this week. 𝕏¹ 𝕏²
• @Beth_Kindig (≈190 likes, ≈25 reposts, ≈26.6k views — the highest-engagement post found in this scope): "Cisco estimates from 2026 to 2035, agentic AI will result in 9X enterprise traffic growth compared to just 2.5X growth without agentic AI." Names $NET, $CSCO, $CRWD as beneficiaries. Verified: those three fell 2.73%, 0.25% and 2.90% respectively on Friday — a long-run traffic thesis diverging from the week's price action. 𝕏
• @saastr (≈5 likes, ≈1,275 views): "Cursor hit a billion dollar run rate in under 2 years, and then 3 months later they got to $2 billion. Top AI companies grew 175% this year." A second post: "Case Study: $100m ARR Gorgias for AI CX" (≈2 likes, ≈1,547 views). Points to AI-native ARR ramps far outrunning classic SaaS. 𝕏¹ 𝕏²

Bearish / cautious

• @HexxRL (same post): Names $ADBE at a 63.8% drawdown and $CRM at roughly -30%, attributing it to seat-decay — the erosion of per-seat pricing by agents. Drawdown figures are the author's; not independently recomputed here. 𝕏
• @StockWorthyApp (≈10 likes): "$ZS Zscaler was down 10.06% on Friday following the sudden resignation of CRO Mike Rich. Investors fear the cybersecurity company may miss on short-term execution." This figure matches the verified snapshot exactly (-10.06%). 𝕏
• @Morpheu5Watcher (≈3 likes): The sharper read — "investors are pricing a risk the company named first. In May, Zscaler blamed departing sales leaders for a slower growth outlook." The market remembers the company's own attribution and won't accept a benign framing now. 𝕏

Chinese software & cloud

• Bearish on legacy ERP vendors · @kxSchoolCN (≈0 likes): "Legacy software companies can't do this — it's not a capability problem, it's a DNA problem. Yonyou and Kingdee can both build finance, inventory and OA software. But open-source their finance software for free? Impossible... it's not that the tech can't, it's that the business model won't allow it." 𝕏
• Bullish on Kingdee · @JKKK_Trader (≈1 like): "$00268 — Kingdee International: high renewal rates on Cangqiong LLM MaaS and the cloud ERP breakeven point... its cloud services segment is expected to reach full profitability for the first time in the second half." Verified: Kingdee (00268.HK) closed 09-28 at HK$7.55, up 2.30%, directionally consistent. 𝕏
• Bullish on domestic compute leasing · @passluo (≈7 likes): "Kingsoft Cloud leased tens of thousands of 5090s in August for a total of RMB 633 million. The 5090 and PRO 6000 aren't actually worse than the Hopper series — they just lack NVLink." Verified: Kingsoft Cloud (03896.HK) closed 09-28 at HK$4.90, down 3.16% — a bullish post against a down day. 𝕏
• Constructive on technology · @shao__meng (≈41 likes): Extensive excerpts from Alibaba Cloud's open-sourced enterprise AI Agent white paper, highlighting that it fills a gap on "engineering, scaling and organizational" dimensions. No bearish framing. 𝕏

4. Buzz & Sentiment Shifts

Oracle / AI infrastructure credit chain: This topic absorbed nearly all of the sector's narrative bandwidth, yet the discussion was high in quality and low in reach — the dozen-plus relevant posts retrieved clustered at 0–7 likes and tens-to-hundreds of views, with no viral breakout. Sentiment was effectively one-sided bearish (no bullish post surfaced on this topic), and the focus escalated quickly from "project delay" to "debt and CDS": @slamwontmiss on the $18B loan and widening CDS, @gin_cryptodiary relaying Morgan Stanley's AI-infrastructure credit warning. Versus last week, this is a new category of bear argument — no longer "is AI demand peaking?" but "can AI supply clear permitting, power and financing?" The tape's reaction stayed measured: ORCL fell just 1.75% on Sept 25 on 23.7m shares and $3.27B of turnover, with no cascading decline.

US SaaS divergence: The infra/security vs. application platforms dichotomy (@HexxRL) was the most structured retail-side framework this week, and Friday's prints partly refuted it — INTU (-0.48%) and ADBE (-1.45%), labelled compressed, outperformed CRWD (-2.90%) and NET (-2.73%), labelled near highs. The real dispersion came from idiosyncratic events, not sector labels: MSFT +3.66% on the Copilot revamp plus the Stifel upgrade, DDOG +4.36% on AI observability demand, ZS -10.06% on a sales-leadership exit. Set against Stratechery's 2026-07-15 description of how a single company's results reset sector sentiment and then spill over, this week ran the other way: the market quarantined single-stock events with no sector-level spillover — worth tracking as a sentiment indicator from here.

Chinese software & cloud: Materially quieter than the US side, and institutional/media coverage was entirely absent — the China whitelists returned nothing in this scope, so all discussion came from long-tail retail accounts (0–41 likes per post). The bull/bear split was roughly balanced: bulls on cloud-ERP breakeven (Kingdee) and tight domestic compute supply (Kingsoft Cloud); bears on the business-model DNA constraint facing legacy ERP. Unlike the US, the dominant Chinese thread remains subscription-transition progress rather than agentic disruption of pricing models — the two markets' narratives are one phase apart. Stratechery's 2026-09-16 analysis of Salesforce deliberately de-emphasising UI as a strategic layer — on the view that UI-based moats are weakening industry-wide — is a mechanism that had not yet entered the Chinese software discussion this week.

5. First-hand Industry Signals (Last 14 Days)

Interpretation of the key events

① Oracle's force majeure: the AI capex constraint moved from chips to power and permitting. Until now the market's bottleneck story sat on compute supply — Stratechery's 2026-01-30 The Chip Fly in the AI Ointment stated plainly that "today we have chip shortages," locating the binding constraint in semiconductors. This week relocated it: Project Jupiter is stuck on a natural gas pipeline permit denied twice by the state, pushing energization from August 2026 to February 2027. The implications differ by party. For Oracle, it shifts schedule risk onto the developer and its financiers by legal means — protecting near-term cash flow while spending counterparty trust. For Blue Owl and similar infrastructure private-credit lenders, it sets a first precedent of a hyperscale offtaker asserting relief, which feeds directly into how they price comparable projects. For neoclouds like CoreWeave, it weakens the premise that long-term contracts equal certain revenue — precisely @stock_duty's "Watch $CRWV sympathy," and CRWV's 2.82% decline that day was directionally consistent. Note that the UBS-attributed "45 builds, $68B stalled or cancelled" is a second-hand relay with no primary note verified and should not be used as an industry aggregate.

② Zscaler down 10% on a CRO exit: pricing power in security SaaS still rests on the sales organization. Three sell-side shops (Guggenheim, Rosenblatt, Mizuho) called it continuity rather than disruption — Mizuho even raised its target to $220, RBC to $236 — yet the stock fell 10.06% in a session. That isn't disagreement so much as the market pricing off a longer memory: @Morpheu5Watcher notes the company itself attributed a slower growth outlook to departing sales leaders back in May. For subscription security broadly, the read-through is that against a backdrop of agentic pressure on per-seat pricing, go-to-market stability is now treated as a standalone risk factor, weighted well above maintained sell-side ratings.

③ Microsoft's product-plus-rating double: the only large-cap software name to meaningfully outperform. The Copilot revamp added a coding tool and a resident agent (@akishore), layered onto Stifel's upgrade to Buy at $575. Stifel's logic deserves its own note: it argues open-weight models actually reinforce Microsoft's LLM-agnostic posture — Azure and Copilot are not bound to a single model supplier, so commoditization at the model layer is a net positive rather than a threat. MSFT closed at $516.17, up 3.66%, on a 28.76x trailing P/E and a $3.83tn market cap. The contrast with Oracle is stark: two hyperscale cloud franchises where Microsoft's AI narrative lands at the software and product layer while Oracle's lands on self-built infrastructure and long-contract backlog — and the market priced those two paths very differently this week.

Disclaimer: This report compiles real-time search results from public X (Twitter) posts alongside public market data. It is provided for informational and educational purposes only, does not constitute a recommendation to buy or sell any security, and does not take into account any individual's financial situation or investment objectives. Third-party views cited (KOLs, media, relayed sell-side research) represent their authors alone and not the views of this platform; items marked "rumor" or "unverified" have not been confirmed against a primary source. Markets carry risk; decisions require independent judgment.