Affirm Holdings (AFRM) climbed 5.5% in premarket trading on Thursday after the fintech company announced the debut of an advanced transformer-based artificial intelligence system engineered to deliver quicker and more precise lending decisions during the checkout process. Affirm is best known for buy now, pay later installment financing — a business where approval decisions have to be returned in the seconds a shopper waits at checkout, putting both speed and accuracy at a premium.

Shares were changing hands near $71.58 ahead of the disclosure — a level substantially below the $98.78 mean price target implied by analyst consensus on the stock.

Nationwide Rollout Built on 14 Years of Proprietary Credit Data

The cutting-edge system has been rolled out nationwide and is now operational across Affirm's U.S. checkout systems, where it powers instant loan approval decisions. The platform leverages 14 years of the company's proprietary transaction-level credit and underwriting information, applying machine learning techniques to evaluate the sequence and timing of financial events in a borrower's credit profile. Affirm characterized the technology as a significant upgrade from its legacy decisioning methodology. Transformer models — the same class of architecture behind modern large language models — are built to recognize how events unfold in sequence, a natural fit for underwriting that weighs the order and timing of a borrower's financial history.

A particularly notable feature of the model is its treatment of consumers with limited credit files. According to Affirm, the AI approved additional candidates who would have faced rejection under the previous framework, including individuals with sparse credit backgrounds and those who lack FICO scores altogether. The gap is a long-standing challenge in consumer credit, where applicants without traditional scores have been difficult for conventional underwriting to assess.

Stronger Performance With Tighter Risk Management

During its initial deployment the system generated 3.4% more finalized transactions relative to a benchmark cohort. Affirm further noted that these additional loans demonstrated superior performance compared with similar expansion efforts under the company's earlier machine learning frameworks. The company said the results indicate that the AI is selecting higher-quality credit risks rather than simply increasing approval volume.

Libor Michalek, President of Affirm, explained that the transformer framework enables the organization to derive fresh insights from datasets it already holds. “Understanding a credit history with greater clarity enables us to responsibly extend approval to more consumers,” he stated.

The system recognizes patterns both within individual credit accounts and across multiple accounts, monitoring how those patterns evolve over time. To keep the model transparent without slowing it down, Affirm developed a specialized algorithm that preserves sufficient speed for real-time checkout processing.

The company emphasized that universal approval is not the objective of the new technology. In Affirm's view, providing credit to individuals who are unable to meet repayment obligations serves no business purpose. With the rollout now live across the country, the 3.4% transaction lift and the early performance of the incremental loans form the baseline against which the model's ongoing credit results can be measured.

Executive Stock Sales Draw Investor Attention

Not every signal pointed upward, however. Chief Accounting Officer Siphelele Jiyane divested 26,980 shares on September 14 at a mean price of $72.19, generating approximately $1.95 million in proceeds from the sale. The transaction decreased his holdings in the company by 12.51%, leaving him with 188,711 remaining shares.

Jiyane had previously sold 25,000 shares on September 4 at $72.41 per share, meaning the two transactions came within roughly ten days of each other. The back-to-back disposals may attract investor attention, though insider sales do not necessarily indicate bearish sentiment.

Latest Quarter Beats on Earnings

On the financial side, Affirm delivered earnings of $4.62 per share in its latest quarter, a figure that significantly exceeded the $0.35 consensus projection from analysts. Revenue reached $1.07 billion, representing 33.1% year-over-year growth, although the top line came in marginally below the $1.11 billion forecast.

Wall Street analysts have generally maintained positive stances on the stock. Piper Sandler elevated its price objective from $103 to $115 alongside an overweight designation, while BMO Capital Markets and Citigroup both confirmed outperform recommendations. Among the 32 analysts monitoring the stock, 24 assign it a Buy rating while 8 recommend Hold, producing a “Moderate Buy” consensus with a mean price target of $98.78.

For technical context, Affirm's 50-day moving average currently stands at $75.43, while the 200-day moving average registers at $66.39 — leaving the premarket price below the shorter-term average but above the longer-term one.