When Wall Street analysts collectively turn to AI: from chips to software, tech investments are being repriced

Wall Street Is Redrawing the AI Industry Map with “Ratings”

This set of Wall Street analyst views compiled by CNBC looks, at first glance, like a普通 stock-rating list: companies such as Nvidia, Apple, Micron, SanDisk, IBM, and Meta all appear on it. But if you place these judgments on the same industry map, you’ll find they point to the same fact — AI is not a single tech subsector, but a force reshaping the pricing logic of the entire technology capital market.

Over the past decade-plus, tech stock valuations have usually revolved around two things: user scale and software subscription revenue. Now, the new key variables are three: computing power, data, and distribution. Whoever controls GPUs, memory, storage, cloud infrastructure, and operating system entry points is closer to the profit pool of the AI era; whoever still relies on the traditional growth narrative must face greater uncertainty.

The common thread in this round of analyst opinions is not a simple reading of near-term earnings, but a response to structural changes in the technology industry.

Computing Power Is No Longer Just a Hardware Business, but the “Basic Energy” of the AI Era

Morgan Stanley remains bullish on Nvidia and, beyond emphasizing its GPU leadership, also sees growth opportunities in its CPU business. The significance of this view lies not only in the fact that Nvidia remains a core beneficiary of AI infrastructure investment, but also in what it says about AI computing evolving from a single accelerator into a more complete system-level platform.

GPUs are still at the center of training and inference, but the real industry trend is moving toward a broader form of heterogeneous computing: CPUs, networking, memory, storage, packaging, and software stacks are all being redefined. In other words, the AI industry is no longer just competition among model companies, but a joint campaign across semiconductors, cloud computing, system architecture, and developer ecosystems.

That is also why memory makers are back in the market spotlight. Morgan Stanley simultaneously raised its price targets for Micron and SanDisk, and judged that the upcycle in storage stocks is not over. What this reflects is not simply a cyclical trade, but a change in the structure of storage demand brought on by AI. Large-model training, vector databases, inference caching, and enterprise data flows are all magnifying the importance of high-bandwidth storage and capacity storage. The bottleneck in the AI era has never been “whether there is a model,” but “whether the model can be deployed into the real world at sufficiently low cost.”

When the market starts chasing memory and storage again, it means capital has already realized: AI infrastructure does not only include the most visible GPUs, but also those quieter, yet equally critical, supply-chain links.

Apple’s AI Narrative Is Shifting from “Feature Upgrades” to “Operating System Rebuilds”Goldman Sachs reiterated its buy rating on Apple and is betting on a next-generation AI Siri on the eve of WWDC. What matters is not any single new feature itself, but that Apple is trying to re-embed AI into devices and the operating system, moving it down from the app layer.

If this move progresses smoothly, Apple’s AI value will not only be reflected in a smarter voice assistant, but in its potential to rewrite the entry logic of human-computer interaction: shifting from tapping apps to task execution driven by natural language; from operating a single app to executing across apps in context; from “users looking for functions” to “the system understanding intent.”

This is highly significant for the entire consumer electronics and platform economy. In the past, the power of the smartphone ecosystem came from app distribution; in the future, AI assistants may become the new middle layer, redistributing search, recommendations, service calls, and commercial traffic. Once Apple finds the balance among on-device processing, cloud inference, and privacy protection, it will not just be filling in Siri’s shortcomings, but competing for the next-generation interaction gateway.

That is also why AI competition will not take place only between OpenAI, Google, and Anthropic. What truly determines commercialization efficiency is who can turn model capabilities into a device-level experience, and then embed that experience stably into billions of endpoints.

Meta’s AI is not an experiment, but the reindustrialization of the advertising machine

Morgan Stanley remains bullish on Meta and links its upside in revenue to AI Search, subscription revenue, core advertising, and a “neocloud back-up plan.” This judgment very typically reflects Meta’s strategic reality: it is not building an abstract AI business, but using AI to further strengthen its advertising distribution machine.

Meta’s core issue has never been “whether it can do AI,” but “whether AI can further improve the efficiency, conversion rate, and monetization density of the advertising system.” If large models can enhance ad understanding, content recommendation, search, and creative generation, then AI will directly affect the quality of ad inventory and the commercial value of each impression.

This means Meta’s AI strategy does not depend on a single product breakout, but is embedded within its entire platform-economy structure. For Meta, AI is not a new standalone business, but a common amplifier for advertising, the social graph, content distribution, and future subscription models.

At the same time, the analysts’ mention of a “neocloud” backup plan also shows that in the AI era, tech giants are unwilling to stake their entire infrastructure on a single supply chain. Compute scarcity, rising costs, and intense competition mean that every major platform is looking for a second or even third supply path. This kind of infrastructure redundancy is becoming the new strategic norm for large tech companies.## IBM’s Value Comes from the “Middle Ground” in the Transition from Legacy Giants to New Architectures

Citi raised its price target on IBM and emphasized that it is participating in two architectural shifts at once: AI and quantum computing. Whether or not the market agrees with this valuation logic, IBM’s case is highly representative: in the AI era, not every winner has to come from the hottest consumer-facing or model-layer plays; some established companies may instead gain new valuation support because they own enterprise customers, hybrid cloud, systems integration, and highly sticky workloads.

IBM matters because it sits in the middle layer of enterprise technology migration. Many large organizations will not move overnight to a purely cloud-native or purely frontier-model stack; what they need is a controlled, compliant, and integrable transformation path. IBM’s opportunity lies here.

More broadly, IBM represents an underrated trend: the commercialization of AI is not happening only in the high-growth stories of consumer internet, but also in the slow rebuilding of enterprise architecture. There is no blockbuster-style growth narrative here, but there is more stable, longer-lasting cash flow and demand for system migration.

From DigitalOcean to Agilysys: Stratification Among Startups Is Intensifying in the AI Era

KeyBanc issued a positive rating on DigitalOcean, and Piper Sandler also initiated coverage of Agilysys. On the surface, these companies do not belong to the most dazzling first tier of AI, but they neatly reveal the structural changes taking place in the startup ecosystem.

In the AI era, startups are no longer simply divided into “model builders” and “application builders”; instead, they are splitting into three layers:

1. Infrastructure companies: Provide cloud, development environments, storage, deployment, and industry-specific platforms. 2. Vertical software companies: Embed AI into specific scenarios such as hospitality, payments, retail, industrials, and healthcare. 3. Distribution companies: Control user entry points, content traffic, or workflow power.

DigitalOcean represents a lightweight cloud opportunity aimed at developers and small and medium-sized businesses; Agilysys is a typical example of vertical software. What they have in common is that AI is not their only selling point, but it may become a key variable in improving product stickiness, customer retention, and workflow automation efficiency.

This is a reality check for the startup ecosystem. The AI era does not inherently reward the companies that are best at telling stories; it rewards those that can embed AI into high-frequency workflows, demonstrate clear ROI, and possess industry distribution capabilities.

Cybersecurity and Infrastructure Substitution Are Becoming the New Capital Risk

Oppenheimer downgraded AT&T, citing among its reasons competitive pressure from SpaceX and low-Earth-orbit satellite constellations. This judgment may appear to concern the telecom industry, but in fact it is tied to the future of AI infrastructure and the digital economy.

The reason is simple: when communications, cloud, endpoints, and compute are all being re-architected, traditional telecom networks are no longer just “connectivity businesses”; they will be challenged by more flexible infrastructure alternatives.The reason is simple: as communication, cloud, devices, and computing power are all being re-architected, traditional telecom networks are no longer just “connectivity businesses” and are instead being challenged by more flexible infrastructure alternatives. Low-Earth-orbit satellites, edge computing, satellite backhaul, remote automation, and industrial interconnection are all changing the value structure of the connectivity layer.

On the other hand, the AI era will also push cybersecurity into a more central position. Model calls, enterprise data access, automated agents executing tasks, and cross-platform permission management all imply a larger attack surface. AI does not just improve productivity; it also increases system complexity. And the more complex a system is, the more important security governance becomes.

In the next few years, the speed at which enterprises purchase AI may increasingly depend on whether they can prove that model calls are auditable, data boundaries are clear, permission controls are traceable, and external dependencies are manageable.

The real change in the industry chain: tech giants are no longer just selling products, but competing for “system positions”

Putting these analyst views together, a deeper trend becomes visible: competition among tech companies is shifting from product competition to competition for system position.

• Nvidia is competing for the position of the computing center.
• Apple is competing for the position of the interaction gateway.
• Meta is competing for the position of content and ad distribution.
• IBM is competing for the position of enterprise architecture transition.
• Micron and SanDisk are competing for the position of storage and data flow.
• DigitalOcean and Agilysys are competing for the position of SMB and vertical-industry workflows.

This means that the winners in the AI era will not necessarily be the companies best at building a single model, but the ones best at embedding models into systems, supply chains, and chains of user behavior. In other words, the endgame of AI is not a single-point product, but an infrastructure order that redistributes technological power.

Conclusion: AI competition is shifting from “who is smarter” to “who is closer to the system core”

The real signal sent by this round of Wall Street analysts is not whether a particular stock rises or falls, but that the market is beginning to reorder companies according to the industrial coordinates of the AI era.

Computing power, storage, operating systems, ad platforms, enterprise software, communications infrastructure, satellite networks, cybersecurity—these formerly dispersed fields are now being reconnected by the same technological revolution. The repricing in capital markets is, in essence, an acknowledgment of one thing: AI is not just a new application layer; it is becoming the underlying organizing principle of the global tech industry.

And when a technology begins to rewrite industrial structure, the most important investment judgment is no longer “Will it take off?” but “Who will it push to the center, and who will it push to the edge?”