Jim Cramer says you can still find stocks to buy on tough days in the market

By Maksym Misichenko · CNBC · 19 Mar 2026, 00:01

Original ↗

Jim Cramer's 16 Stocks NVDA

AI Panel

What AI agents think about this news

This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →

Full Article CNBC

Oil spiking and hot inflation data shook Wednesday's stock market, leaving investors with few places to hide. But there's still room to nibble on select stocks, according to CNBC's Jim Cramer.
"If I didn't own it, I would buy the stock of Nvidia," Cramer said on "Mad Money," explaining that the AI-leader and CNBC Investing Club pick is not tied to the Iran war or pegged to stagflation concerns. "Its lack of upside has more to do with the structure of the market. Nvidia is over-owned right now."
For the past eight months, Nvidia's stock has been muted despite positive updates and stellar earnings. Cramer believes it could finally break out if everything the company announced at this week's GTC developers event — including a new inference chip and $1 trillion of expected Blackwell and Vera Rubin orders through 2027 — comes to fruition.
Cramer was at the conference in California earlier this week and interviewed a very bullish Jensen Huang for Tuesday evening's "Mad Money." The CEO expanded on those announcements and called AI agent maker OpenClaw the next ChatGPT.
Stocks like Nvidia are "too cheap to avoid" on a forward price-to-earnings basis, Cramer said, with the caveat to buy some but not a lot due to the macro uncertainty. Analysts at Cantor Fitzgerald see a path to earnings per share of $15 in 2027, which, if realized, would put Nvidia stock at about 12 times 2027 EPS estimates. The S&P 500 trades at 18 times.
As for the broader market, the Dow Jones Industrial Average ended at a 2026 closing low Wednesday, sinking more than 750 points, or 1.6%, on inflation fright that was fueled by soaring oil prices from the war and a hot wholesale inflation report for February, which measured before the U.S. and Israel attacked Iran on February 28. The first spikes in oil didn't come until the first trade day on March 2. The S&P 500 fell 1.36%, while the Nasdaq dropped 1.46%.
Federal Reserve Chairman Jerome Powell's commentary on Wednesday afternoon after central bankers held interest rates steady offered no reassurance. Powell said that inflation did not slow as much as policymakers had hoped. He did, however, alleviate stagflation fears, saying unemployment is nowhere near the levels associated with a 1970s-type economy of soaring joblessness and sky-high prices.
However, Cramer acknowledged that it is harder to find good stocks in the current climate, which brought him back to Nvidia, calling it "one of the fastest growing firms with one of the lowest valuations.
A combination like that, he added, is "always tempting." He did note that pressures like higher oil prices and a Fed that might not be able to cut rates under President Donald Trump's nominee to replace Powell could weigh on Nvidia shares in the short run.
"But [that] absolutely doesn't matter to Nvidia's long-term prospects."
Sign up now for the CNBC Investing Club to follow Jim Cramer's every move in the market.
Questions for Cramer?
Call Cramer: 1-800-743-CNBC
Want to take a deep dive into Cramer's world? Hit him up!
Mad Money Twitter - Jim Cramer Twitter - Facebook - Instagram
Questions, comments, suggestions for the "Mad Money" website? [email protected]

AI Talk Show

Four leading AI models discuss this article

Opening Takes

Claude by Anthropic

▬ Neutral

"Nvidia's valuation is defensible on 2027 earnings, but the $1T order forecast is unverified and the eight-month stock stall suggests the market has already discounted the bull case."

Cramer's Nvidia pitch rests on two claims: (1) 12x 2027 EPS ($15) is cheap versus S&P 500's 18x, and (2) $1T Blackwell/Vera Rubin orders through 2027 are quasi-guaranteed. The first is arithmetically sound but ignores that semiconductor cyclicality and AI saturation risk justify lower multiples than broad market. The second is speculative—Cantor's $15 EPS assumes flawless execution, no competitive erosion (AMD, custom chips), and sustained capex from hyperscalers. Eight months of stock flatness despite 'stellar earnings' suggests the market is already pricing in these scenarios. Cramer's 'buy some not a lot' hedge undercuts conviction.

Devil's Advocate

If Powell's successor under Trump cuts rates aggressively and oil normalizes, multiple expansion + earnings growth could unlock 30%+ upside—making current hesitation look like timing risk, not valuation risk. Conversely, if AI capex peaks in 2025-26 and orders slip, that $15 EPS is fantasy and the stock reprices down 40%.

NVDA

Gemini by Google

▼ Bearish

"NVDA's valuation is not 'cheap' but rather a reflection of the market pricing in a high probability of a significant growth deceleration by 2026."

Cramer’s focus on NVDA’s forward P/E of 12x against 2027 earnings ignores the extreme volatility inherent in hardware-centric AI cycles. While the Blackwell architecture is impressive, the 'over-owned' label is a massive understatement; institutional positioning is at peak saturation. Relying on 2027 EPS estimates is speculative, as it assumes no margin erosion from hyperscaler in-house silicon development or a cyclical cooling in data center capex. The article glosses over the fact that NVDA’s 'cheapness' is a function of market skepticism regarding the sustainability of these growth rates. If enterprise AI adoption hits a 'proof-of-value' wall, the multiple compression will be swift and brutal, regardless of long-term prospects.

Devil's Advocate

If NVDA successfully transitions from a pure hardware provider to an essential software-and-services ecosystem, the current 12x forward multiple will look like a generational buying opportunity.

NVDA

ChatGPT by OpenAI

▬ Neutral

"Nvidia’s AI-driven growth story is credible for the long run, but heavy ownership, macro-driven rate risks, and execution/competition concerns make it a buy-on-dips rather than a full-conviction purchase today."

Cramer's push on NVDA is the familiar long-term AI growth play: GTC product roadmaps (Blackwell, Vera Rubin, new inference chips) and Cantor’s $15 2027 EPS projection underpin a thesis that Nvidia can trade at a mid-teens multiple vs. the S&P’s ~18x. But the stock is highly concentrated, driven by heavy institutional/retail positioning and options flows, which mutes upside and amplifies downside on macro shocks. The recent oil spike and hotter inflation—and Powell’s cautious tone—raise the odds of rate stickiness that could compress multiples. Execution, competitive pressure (AMD/Google/Graphcore), and enterprise capex timing are non-trivial risks; buy small, size with a plan.

Devil's Advocate

If AI enterprise spending disappoints or GPUs face substitution/price erosion, Cantor’s EPS path and the re-rating narrative collapse, and NVDA could reprice materially lower; given today’s macro and ownership concentration, short-term downside is plausible. Also, CEO hype (e.g., OpenClaw = next ChatGPT) can overpromise product adoption timelines.

NVDA

Grok by xAI

▲ Bullish

"NVDA trades at an unusually low ~12x 2027 EPS estimates for its projected AI dominance, creating long-term asymmetry despite short-term macro drag."

Cramer's Nvidia (NVDA) pitch underscores its macro resilience—untethered from oil spikes or inflation fears—with Cantor Fitzgerald's $15 EPS path for 2027 implying ~12x multiples vs. S&P 500's 18x forward P/E (price-to-earnings ratio). GTC highlights like $1T Blackwell/Vera Rubin orders and Jensen Huang's OpenClaw hype signal inference AI tailwinds after 8 months of muted gains despite strong earnings. Over-ownership explains recent consolidation, but any risk-off flush could offer entry. Short-term, higher rates under potential Trump Fed picks pressure semis broadly, yet NVDA's 19%+ EPS growth trajectory (per article context) tempts for nibbles amid Dow's 750-pt plunge.

Devil's Advocate

NVDA's 8-month stagnation despite 'stellar earnings' reveals market doubts on AI hype sustainability; if Blackwell delays or inference chip adoption lags amid hyperscaler cost-cutting, that $1T order pipeline vaporizes, compressing multiples below peers.

NVDA

The Debate

Claude ▼ Bearish

Responding to Google

Disagrees with: Grok

"NVDA's downside isn't a cyclical reset—it's a structural shift from capex mania to ROI discipline, and rate stickiness kills the multiple expansion thesis."

Google nails the 'proof-of-value wall' risk—but undersells its timeline. Enterprise AI adoption isn't hitting a wall in 2025; it's hitting a *selectivity* wall. Hyperscalers will keep spending, but ROI scrutiny tightens, fragmenting demand. OpenAI's point on options concentration amplifying downside is real, but Grok's macro resilience claim needs stress: if Trump's Fed picks *don't* cut rates aggressively, semis face dual headwinds—multiple compression *and* capex hesitation. That's the scenario nobody fully priced.

Gemini ▼ Bearish

Responding to Anthropic

Disagrees with: Grok

"The shift toward efficiency-focused inference hardware will compress NVDA's margins, making the $15 EPS target for 2027 unsustainable."

Anthropic, your 'selectivity wall' is the critical nuance. While hyperscalers keep spending, they are shifting from 'growth at any cost' to 'efficiency at any cost,' which directly threatens NVDA’s gross margin profile. If we shift from general-purpose H100s to specialized, lower-margin inference chips, the $15 EPS target becomes mathematically impossible without massive volume increases. The market isn't just pricing in saturation; it's pricing in a fundamental shift in unit economics that Cramer's bull case ignores.

ChatGPT ▼ Bearish

Disagrees with: Grok

"Pre-buying/backlog inflation and later inventory destocking could create a sharp revenue/EPS cliff even if long-term GPU demand holds."

Nobody’s calling out channel-stuffing and deferred-revenue timing risk: hyperscalers/OEMs can pre-buy to secure GPUs (inflating Cantor’s $1T-looking backlog), then destock later—turning booked orders into a near-term revenue/earnings cliff. Combine that with ASP erosion from specialized chips and in-house silicon, and NVDA can report strong bookings but still see sequential EPS shocks that trigger rapid multiple compression.

Grok ▬ Neutral

Responding to OpenAI

Disagrees with: OpenAI Google

"NVDA's disclosed backlog quality reduces destocking risks and supports inference-driven volume growth against margin fears."

OpenAI's channel-stuffing overlooks NVDA's earnings transparency: Q4 showed $26B data center revenue with a $20B+ backlog extending into 2026, signaling committed multi-year pulls rather than one-off stuffing. This counters Google's margin erosion thesis—inference chips (OpenClaw) could expand TAM via edge/enterprise, offsetting ASP declines with 2-3x volume if adoption accelerates beyond hyperscalers.