Salesforce raised its full-year revenue and profit forecasts on Wednesday and deepened its partnership with Anthropic, giving investors fresh evidence that the rise of generative AI may be creating a new growth engine for the enterprise software giant rather than simply threatening its traditional business.

Shares of Salesforce jumped 14% in extended trading after the company unveiled “Claudeforce,” a new initiative that brings Anthropic’s Claude AI models together with Salesforce’s customer data, applications and workflows. The agreement expands a partnership announced in June and comes as investors have been questioning whether capable AI models could erode the value of conventional software.

The results provide a more complicated picture. Salesforce’s quarter ended July 31 produced revenue of $11.35 billion, up 11% from a year earlier, while the company raised its fiscal 2027 revenue forecast to between $46.1 billion and $46.4 billion from a previous range of $45.9 billion to $46.2 billion. It also lifted its adjusted earnings forecast to $16.67-$16.71 per share from $14.06-$14.12.

The stronger outlook has drawn interest because Salesforce is trying to prove that AI agents can become a meaningful source of incremental revenue while protecting the value of its core customer relationship management business.

Salesforce said momentum in Agentforce, Data 360 and Slack is helping offset volatility in traditional license revenue. The company is also expecting contributions from its planned acquisitions of Contentful and Fin, announced in June.

Agentforce is at the center of the strategy. Salesforce has been moving beyond AI assistants that simply answer questions toward autonomous agents capable of carrying out tasks in sales, customer service and other business functions. That potentially changes the economics of enterprise software because customers could pay not only for software seats, but for AI-driven work performed on their behalf.

Recent results suggest that business is gaining scale. Agentforce and Data 360 together generated nearly $3.9 billion in annual recurring revenue, according to MarketWatch, while Salesforce said it processed more than 7 billion agentic work units, including 3.2 billion during the latest quarter.

Besides concern over AI disruption, the threat to Salesforce has been that customers could use frontier models from companies such as Anthropic and OpenAI to build their own applications, reducing their dependence on large enterprise software vendors. Salesforce is instead attempting to make those models part of its own platform.

“Claudeforce” is therefore not just a conventional technology partnership but a representation of a strategy in which Salesforce provides the enterprise data, business processes and software environment while Anthropic supplies a powerful general-purpose AI model. The arrangement could allow Salesforce to benefit from improvements in frontier AI without having to develop every underlying model itself.

It also shows that model developers increasingly need access to proprietary business data and established distribution channels, while software companies need access to sophisticated AI models. Partnerships such as Salesforce-Anthropic can allow both sides to capture value from that relationship.

The financial results, however, require some qualification. Salesforce’s adjusted earnings of $5.90 per share were more than double the year-earlier level, but $2.53 per share came from gains on strategic investments, including its investment in Anthropic. Share repurchases also reduced the number of outstanding shares and boosted per-share earnings. Excluding the investment gain, adjusted EPS was $3.37, according to MarketWatch.

That makes the revenue outlook and the underlying operating performance more important indicators of whether Salesforce has genuinely turned the corner.

The company is also facing a broader challenge in its traditional software business. AI can make some software functions easier to reproduce now, potentially putting pressure on license growth and pricing. Salesforce’s answer is to move its value proposition higher up the stack, from providing software used by employees to providing an operating layer where employees and AI agents work together.

That approach helps explain why the company has been so aggressive in promoting Agentforce. If AI agents become the primary interface through which employees interact with corporate systems, Salesforce wants those agents running on its platform and using its data rather than bypassing it.

The results offer some support for that argument, but the durability of the trend remains the key question. Analysts have noted that Salesforce must demonstrate that AI-related growth is not simply shifting existing spending from conventional software products into new AI offerings. It needs to show that customers are expanding their overall spending because AI agents are producing measurable productivity gains.

The partnership with Anthropic also highlights a potential tension. Salesforce is relying on an outside model provider even as it seeks to make its own platform indispensable. The company now needs to maintain control over customer data, workflows, and distribution while avoiding excessive dependence on any single AI model provider.

Still, the immediate market reaction suggests investors were encouraged by the combination of stronger guidance and accelerating AI adoption. Salesforce had been caught in the selloff in software stocks as investors questioned whether AI would make established applications obsolete. However, the latest results provide a counterargument: rather than replacing enterprise software outright, AI may increase the value of platforms that already control business data and workflows.

SK Hynix said on Thursday it plans to begin volume production of its next-generation HBM4E memory chips at its Indiana facility in the third quarter of 2029, as the South Korean chipmaker prepares for what its chief executive expects to be a prolonged shortage of advanced memory driven by the expansion of artificial intelligence.

Chief Executive Kwak Noh-Jung said the company expects the memory shortage to persist through the end of 2030, offering a bullish outlook for a market that has emerged from a severe downturn and become one of the biggest beneficiaries of the AI infrastructure buildout.

“We see no clear signs of a memory downturn,” Kwak said, adding that any eventual slowdown was more likely to involve a moderation in demand than a sharp decline.

The comments underline how fundamentally AI has changed the memory-chip market. HBM, or high-bandwidth memory, has become an essential component of advanced AI accelerators because it allows processors to access large amounts of data at very high speeds while using less power than conventional memory configurations.

SK Hynix is investing more than $4 billion in its Indiana project, marking a major expansion of its U.S. manufacturing footprint and bringing advanced packaging and research capabilities closer to some of its largest customers, including Nvidia, Microsoft and Alphabet’s Google.

The facility at Purdue Research Park in West Lafayette is expected to begin cleanroom operations in the second half of 2028, with volume production of HBM4E scheduled for the third quarter of 2029.

Kwak said the site could eventually reach annual production capacity of hundreds of thousands of wafers. By 2030, SK Hynix expects Indiana to become an important U.S. production base for HBM as the company builds what it describes as an advanced memory supply chain in the country.

The facility will include a next-generation HBM packaging line and an AI semiconductor research and development center. That combination is strategically important because HBM production involves more than conventional wafer manufacturing. Advanced packaging is increasingly a critical part of AI-chip performance, as memory must be closely integrated with processors to meet the bandwidth and power requirements of large AI systems.

The investment also gives SK Hynix a way to diversify production geographically. Much of the world’s semiconductor manufacturing remains concentrated in Asia, while U.S. technology companies are rapidly expanding data-center capacity domestically. Locating advanced memory packaging and research in the United States could reduce logistical constraints and improve supply-chain coordination with major customers.

The timing reflects the extraordinary growth in demand for AI computing. Nvidia forecast on Wednesday that its revenue would increase 70% in the next fiscal year, reinforcing expectations that spending on AI data centers will remain elevated.

For SK Hynix, that demand has transformed HBM from a relatively specialized memory product into a major profit driver. SK Hynix held 58% of the global HBM market by revenue in the first quarter of 2026, according to Counterpoint Research, compared with 21% each for Samsung Electronics and Micron Technology. Its leading position gives the company an important advantage as AI chipmakers compete for increasingly scarce advanced memory.

HBM is also structurally different from traditional memory products. The chips are vertically stacked and tightly integrated with AI processors, requiring sophisticated manufacturing and packaging capabilities. Those technical barriers make it harder for new competitors to enter the market and have allowed leading suppliers to command stronger pricing than they typically achieve in commoditized memory markets.

That dynamic is helping explain Kwak’s unusually long-range outlook. Memory manufacturers have historically endured sharp boom-and-bust cycles, with periods of oversupply causing prices and profits to collapse. The rapid expansion of AI infrastructure, however, has introduced a new source of demand that is less dependent on traditional PCs and smartphones.

SK Hynix is effectively betting that AI accelerators, data centers and increasingly sophisticated computing systems will keep absorbing HBM output even if other segments of the semiconductor market weaken.

The company is not relying solely on its Indiana investment. Earlier this month, its board approved about 54.3 trillion won ($38.3 billion) in investments through 2031, including 35.2 trillion won for the second phase of its semiconductor manufacturing facilities in South Korea.

The U.S. project is also supported by Washington’s efforts to build domestic semiconductor capacity. The U.S. government finalized $458 million in CHIPS Act grants and up to $500 million in loans for SK Hynix in December 2024.

The Indiana facility is expected to generate about 7,000 direct and indirect jobs locally and strengthen the U.S. semiconductor supply chain.

SK Hynix, however, has been aiming to capture a large share of the AI industry; the company is positioning itself close to the center of the AI hardware ecosystem at a time when access to advanced memory has become one of the key constraints on scaling AI systems.

The company’s lead in HBM also puts it in direct competition with Samsung and Micron, both of which are investing heavily to expand their share of the market. The three memory giants are now competing not simply on production volumes but on the ability to deliver more advanced generations of HBM with higher bandwidth, lower power consumption and improved integration with AI processors.

That competition could intensify as customers develop increasingly powerful AI accelerators. SK Hynix’s decision to commit to HBM4E production years in advance signals that the company expects technological leadership and manufacturing capacity to remain critical even after today’s AI chips are replaced by newer generations.

Kwak’s forecast that shortages could persist until the end of 2030 is therefore more than a near-term pricing call. It is a bet that AI will fundamentally reshape the memory cycle and sustain demand for advanced products well beyond the current wave of data-center construction.

The risk, as always in semiconductors, is that companies collectively build capacity faster than demand develops. But SK Hynix’s investment strategy suggests it currently sees the greater danger in being unable to supply the next generation of AI systems than in having too much advanced memory capacity.