In the long history of capital markets, every tech bull market has followed a similar script:

It begins with imagination fueled by technological breakthroughs, thrives on capital-driven narratives, and ultimately faces the same harsh reality—the narrative must be validated by earnings.

At the peak of the AI hype, after all the grand visions of Artificial General Intelligence (AGI) have been repeatedly touted, the market’s only real concern is whether the money has actually been made.

Passion fades, and rationality returns. Investors are no longer satisfied with listening to tech companies paint grand visions of the future. They begin to scrutinize balance sheets and calculate return on investment.

This time, while global investors focus on NVIDIA, the earnings of two other companies are seen as deeper 'stress tests': Micron and Oracle.

Why them? Because they occupy hidden yet critical corners of the AI value chain, holding the most authentic supply and demand data.

Their performance will directly determine whether the AI bull market continues to surge or undergoes a profound valuation correction.

Key Tests for the AI Bull Market:

Memory Prices and Cloud Orders

The 2026 AI rally is built on two core assumptions: sustained explosion in computing power demand and expanding corporate AI spending. These assumptions seem self-evident, but in capital markets, what seems obvious is often the most dangerous trap.

If this logic holds, the most direct evidence will not come from unprofitable model companies but from two types of infrastructure firms: memory chip manufacturers and AI cloud service providers.

The former represents the authenticity of hardware demand for computing power, while the latter represents the certainty of corporate AI adoption. This is why the market is focusing on Micron and Oracle this earnings season.

From an industrial standpoint, these two companies occupy two critical nodes in the AI value chain, acting as sentinels guarding the entrance to the AI economy.

Micron operates at the computing hardware layer. In AI servers, while GPUs are the brain, DRAM and NAND are the memory and storage—indispensable core components for GPU servers. Without high-speed storage, computing power cannot be unleashed.

Oracle, meanwhile, operates at the AI infrastructure layer.

Its OCI cloud platform is becoming a key carrier for large-scale AI training clusters, especially for enterprises seeking alternatives to NVIDIA GPUs. Oracle is a major supplier of computing power.

As AI investment enters a phase of massive capital expenditure, these two companies will serve as the earliest 'thermometers' of demand changes. Hardware manufacturers will feel order fluctuations first, while cloud providers will detect computing power shortages earliest.

If both companies continue to exceed earnings expectations, the market will view AI demand as still strong, justifying current lofty valuations. If earnings disappoint, the valuation logic for tech stocks may face re-examination, with investors questioning: Is AI all thunder and no rain?

In other words, these two earnings reports belong not just to the companies themselves but to the entire AI rally. They mark a critical milestone in AI’s transition from 'concept validation' to 'commercial validation.'

Micron’s Key Variable: A Potential

DRAM Super Cycle Reminiscent of the 1990s

In the AI supply chain, memory chips have become the most undervalued segment. The market focuses too much on GPU computing power while overlooking the data costs behind its advancement.

Improvements in GPU computing power mean exponential growth in data throughput, with DRAM and NAND serving as the core infrastructure for data flow. The stronger the computing power, the more insane the demand for memory bandwidth and capacity becomes.

Citi’s latest forecast offers a striking figure: by 2026, DRAM average selling prices could rise 171% year-over-year, while NAND could climb 127%. If this trend holds, the memory industry may be entering an extremely rare cycle.

The last similar scenario was the DRAM cycle during the Windows PC boom of the 1990s. Back then, surging PC shipments outstripped DRAM supply, leading to years of rising memory prices and a golden era for storage giants.

Today, AI servers may be replicating this structure. An AI server equipped with high-end GPUs requires far more DRAM than traditional servers, especially high-bandwidth memory (HBM) and other premium storage, which are in short supply.

HBM has become standard for AI chips, but its production capacity is limited, and its pricing is highly elastic. Meanwhile, after years of industry downturns, global memory manufacturers are extremely cautious about investing in new wafer fabs. The restraint on the supply side and the explosion on the demand side have created a perfect scissors gap.

This combination of surging demand and supply restraint suggests the memory industry may enter an extended cycle. This is why multiple institutions are continuously raising their target prices for Micron—Citi lifted its target to $430, while Susquehanna offered an aggressive $525 forecast.

Behind these numbers lies capital’s bet on a memory super cycle.

For the market, Micron’s earnings report matters not just for its profit figures but for answering a bigger question: Is AI creating a new memory super cycle?

If Micron’s guidance is strong, it means AI hardware demand is not only real but accelerating. If guidance is weak, it may suggest downstream clients are digesting inventory, and AI hardware investment growth is slowing.

Oracle’s Real Challenge:

Can AI Orders Outpace Cash Flow Pressures?

While Micron faces cyclical challenges, Oracle confronts a different issue: the financial cost of AI growth. Over the past year, Oracle has emerged as a dark horse in the AI cloud race.

Amid competition from Amazon, Microsoft, and Google, Oracle’s OCI cloud business achieved 68% revenue growth last quarter, with remaining performance obligations (RPOs) reaching a staggering $523 billion.

These numbers indicate a fact—AI companies are booking massive computing resources, and Oracle is becoming a key platform for them.

The market has even heard of multiple high-profile collaborations, including a long-term cloud agreement with Meta and a critical role in OpenAI’s Stargate project. Oracle seems to have found a second growth engine.

However, this is precisely where the problem lies. To fulfill these orders, Oracle is undergoing extremely aggressive capital expenditure expansion. The company has raised its annual capex from $35 billion to nearly $50 billion.

This has directly led to a result: Oracle’s financial model has begun to fluctuate sharply. Over the past 12 months, the company generated about $22.3 billion in operating cash flow but spent $35.5 billion on capex, leaving free cash flow at -$13.2 billion.

This means Oracle is experiencing a typical 'AI infrastructure bloodletting period.' For a mature software company, negative free cash flow is a dangerous signal.

For investors, the key question is no longer order size but whether these AI orders can convert into revenue and cash flow quickly enough.

Cloud infrastructure has long construction cycles and slow returns. If revenue recognition cannot keep pace with capex, the company’s financial health will suffer.

If OCI growth remains high and order fulfillment speeds up, the market may re-embrace this 'capital-intensive AI cloud' valuation logic, treating it as a high-barrier business similar to telecom infrastructure.

But if capex continues to expand while revenue realization lags, Oracle could find itself in an awkward position—transforming from a high-margin software company into a high-risk AI infrastructure firm.

This identity shift will directly impact its valuation multiples.

Conclusion:

Tech Stock Rebound Requires More Than Just Stories

After two years of AI mania, the market is gradually entering a more realistic phase. Investors are realizing that AI is not just a technological revolution but a capital-intensive industrial competition.

GPUs require capital, data centers require capital, and cloud infrastructure requires capital. All technological progress must ultimately reflect on the bottom line of financial statements.

Thus, whether this tech stock rally can continue largely depends on two questions: Is AI demand truly strong enough? And can this demand quickly convert into cash flow?

Micron and Oracle’s earnings reports stand at the intersection of these two questions. One represents the intensity of hardware demand, while the other represents the efficiency of commercial adoption.

If memory prices continue to surge and AI cloud orders begin to accelerate, tech stocks could see a new rebound, proving that beneath the AI bubble lies a solid economic foundation.

But if earnings reveal slower-than-expected demand realization, the AI sector will face a more serious valuation re-rating, with companies supported solely by narratives at great risk.

For investors, these two earnings reports matter far beyond the companies themselves. They serve as a test—of whether the underlying logic of the AI bull market still holds.

At this critical juncture, staying sober, focusing on data rather than noise, is the only path to navigating the cycle.

A tech stock rebound requires not just stories but hard-earned earnings validation.