IBM's $240M AI Cloud Bet Offers a Counterweight to a Bruised Mainframe Franchise

IBM secures $240M AI cluster deal with Together AI, but faces securities probe and 22% YTD stock decline.

The arithmetic of IBM's current predicament is stark. Mainframe revenue collapsed 42 percent in July, yet management held firm to its 2026 constant-currency growth forecast of 4 to 5 percent. Skeptics called that math optimistic. A freshly signed deal with Together AI now provides at least one concrete data point in the company's favor.

The multi-year agreement, valued at $240 million, will see IBM operate a large-scale inference cluster built on NVIDIA HGX B300 technology within its own cloud, with services slated to go live in the first quarter of 2027. It is a meaningful validation for the hybrid-cloud arm of the business, which has labored in the shadow of the flagging mainframe unit.

A Legal Cloud Hangs Over the Recovery

That the company needs such wins is beyond dispute. On Friday, the law firm Bleichmar Fonti & Auld announced it was opening an investigation into potential securities fraud claims against IBM. The probe centers on possible misrepresentations regarding the outlook for the Z-series product line and the pace of new business closures, triggered by a single-day share price collapse of 25 percent in July.

The stock has clawed back some ground since then—up 2.6 percent over seven days and 4.4 percent over 30—but the scars remain visible. At Friday's close of €203.30, down 0.8 percent on the session, the shares are still 22 percent below where they started the year and a full 31 percent adrift of the 52-week high of €292.85.

A Multi-Pronged Technology Offensive

The Together AI contract is hardly an isolated gambit. At last week's Hot Chips conference, IBM unveiled its first dual-architecture mainframe processor, capable of running IBM Z and Arm-native Linux workloads simultaneously. Built on a two-nanometer process, the chip allows Arm-native AI applications to execute directly on mainframes without virtualization—a pitch aimed squarely at heavily regulated sectors like banking and defense, where data sovereignty is paramount.

The quantum division also logged progress. IBM reported a milestone on its Starling roadmap: two modular cryogenic systems successfully connected and cooled within a shared environment. This follows the recently completed acquisition of HRL Laboratories from Boeing and General Motors, which brings silicon-spin qubit and quantum sensing expertise into IBM's hardware pipeline.

Sales reach is expanding too. TD Synnex has extended IBM's distribution network into 20 additional countries across Europe, Asia-Pacific, and Latin America under the "Select Blue" partner program. And on the consumer-facing front, a partnership with the USTA will bring AI-powered fan features—including a "Serve Quality" metric—to the US Open in 2026.

Insider Activity and Institutional Divergence

Investors tracking insider behavior will note that Senior Vice President Robert David Thomas sold 25,000 shares at an average price of $230.32, trimming his direct stake by 34.34 percent to 47,800 shares. Institutional moves have been mixed: Welch & Forbes LLC increased its position by 30.4 percent during the second quarter, while Pictet Asset Management Holding reduced its holdings by 4.7 percent.

Shareholders on record as of August 10 will receive a quarterly dividend of $1.69 per share on September 10—a measure of predictability amid the share price turbulence and the ongoing legal review.

The Broader Investment Case

IBM's positioning stands in sharp contrast to the aggressive growth narrative of rival Oracle. While IBM has morphed into a consulting and hybrid-cloud specialist—with roughly 30 percent of its consulting backlog now tied to generative AI projects—Oracle has become the platform of choice for massive AI training workloads, landing marquee contracts with OpenAI and the US Department of Defense.

The financial profiles could hardly diverge more. IBM trades at a forward P/E of 19.6x with a 2.5 percent dividend yield and positive free cash flow of $12.2 billion over the trailing twelve months. Oracle commands a 23.0x multiple, yields just 1.0 percent, and burns through cash at a rate of $23.7 billion. S&P has downgraded Oracle's credit rating to BBB- as long-term debt ballooned past $122 billion to finance GPU clusters, while IBM retains a solid A- rating and a 31-year streak of dividend increases.

IBM's strategy is one of embedded AI rather than raw scale. Its Granite models now sit on SAP platforms, and the company is expanding its Sovereign Core catalog to let governments run AI workloads without surrendering data boundaries. The upcoming rollout of the watsonx HSM Manager underscores a focus on high-margin security and compliance software rather than commodity compute volume.

Oracle, by contrast, is betting everything on insatiable AI demand. Management projects $90 billion in total revenue for fiscal 2027, fueled by 20 new cloud regions and eight high-density AI superclusters. The company's $638 billion backlog and ability to ink $30 billion annual contracts with unnamed partners testify to demand that few can match—but the balance sheet leaves little room for error.

A Tale of Two Charts

Technically, IBM presents the calmer picture. The stock has recovered from its July lows and trades near its 50-day moving average, with support at $215 and resistance at $245. The relative strength index sits at a neutral 52, suggesting room for further gains if the next quarterly report confirms expanding software margins.

Oracle's chart tells a more volatile story. After failing to break through resistance at $150, the shares entered consolidation, with a 3.16 percent pullback stalling momentum. A break below $142 could open the door to a test of the $125 zone, where institutional buying was strong earlier this year.

For investors, the choice ultimately comes down to temperament. IBM offers a defensive tech position with a generous dividend and a unique hardware ecosystem that locks in regulated-industry customers. Oracle is the high-risk alternative for those convinced the infrastructure phase of AI is only beginning. The question is whether Oracle can convert its enormous backlog into genuine cash flow before the cost of its debt-fueled expansion becomes unsustainable—and whether IBM's quieter, more diversified path can deliver the growth its guidance promises.