标普维持甲骨文的投资级评级，但数据并不支撑这一评级结论
人工智能企业能否创造可观投资回报，目前仍存在巨大不确定性。
凯文·科哈基，CAE咨询公司创始人、普渡大学会计学副教授。图片来源：Courtesy of CAE Consulting
如果我告诉你，一家企业未来数年营收预计增长240%、债务飙升410%，几乎无法产生正向现金流，而债权投资者依旧不改对公司的评级，你会相信吗？如果我再告诉你，这家企业正大举投入一项尚未产生足够投资回报的技术，试图实现转型，你会更有信心吗？我也不会。
这就是甲骨文当前的人工智能投资叙事，以及标普全球评级（S&P）给出的评级。
三大评级机构（标普、穆迪、惠誉）为各类组织提供信用评级。投资者据此评估企业的融资成本。各机构采用相近的评级体系，衡量债权投资者无法收回出借资金的概率。其中，低风险企业获得投资级评级（AAA、AA、A、BBB），高风险企业则获得投机级评级（BB、B、CCC、CC、C、D）。
简言之，企业评级越低，融资成本越高。债务投资者只能收回本金及利息，没有额外收益空间。因此他们更关注资本回收率，而非资本回报率。
在研读标普2026年7月9日将甲骨文信用评级下调至BBB-（投资级评级中的最低等级）的公告后，我心中的疑问远多于答案。如前所述，标普预测2022财年至2028财年甲骨文营收将增长239%，债务与现金流亦将增长400%–450%。然而遗憾的是，在持续布局人工智能业务后，其实际现金利润（如自由现金流）预计反而下降32%，且2025至2027年间将持续为负。
标普在2026年7月13日的“甲骨文评级下调解读”电话会议中指出，甲骨文和SpaceX“无疑都是投资级评级中的特例”。标普进一步指出，尽管甲骨文目前的信用指标尚未达到投资级标准，但“维持其投资级评级的原因在于，我们认为随着人工智能业务规模扩大，甲骨文能在合同的第三、第四、第五年实现现金回流。因此，我们仍给予企业时间验证商业模式，在此期间，我们将获得更多数据点，对甲骨文人工智能业务前景更有信心”。
需要说明的是，我并不反对评级机构给予企业时间验证商业模式，但相关数据必须经得起推敲。
出乎意料的是，2022至2028年间，甲骨文债务规模预计增长412%，而利息支出仅增长271%。只有债务利率大幅下行，才会出现这种情况。若甲骨文财务状况持续恶化，这种情况不该出现。信用违约互换（CDS）市场也印证了这一点：5年期CDS合约利差近期超过200个基点，上一次达到该水平还是2008年全球金融危机期间。
倘若标普对甲骨文的预测——尤其是2027年之后的预测——充满信心，那么结论将更具说服力，但事实并非如此。标普坦言：“甲骨文现已进入‘用业绩说话’的阶段，前景不明朗，且仍存诸多疑问。”
这种不确定性部分源于甲骨文的商业模式转型及其与OpenAI的关系。标普将这一新商业模式定性为重资产模式，认为其属于“缺乏护城河的业务”，即不具备核心竞争优势。既然如此，标普为何一边预期甲骨文在2028年前实现营收与会计利润大幅增长，另一边却警示其“盈利路径存在不确定性”？
这种不确定性也源于很难预测甲骨文为实现其人工智能雄心所需投入的资金规模。
具体而言，在与甲骨文沟通后，标普不得不将2027年资本支出指引从600亿美元上调至950亿美元，涨幅近60%。标普坦言对此感到无奈，称其需要“不断追赶”持续攀升的资本开支预测。谁不是呢？
标普当前的信用评级与稳定展望，基于两个前提：一是甲骨文将竭力维持投资级评级，二是公司未来可能通过增发股权稳定资产负债表。标普提示，如果甲骨文无法维持或降低当前债务/息税折旧摊销前利润比率，或是未能在2028年产生正向自由现金流，其评级恐将面临下调压力。颇具讽刺意味的是，从当下到2028年，甲骨文预计只有2028年能实现正向自由现金流。正如我们之前讨论的，要达成这一目标，诸多环节都必须完全符合预期。
人工智能企业能否创造可观投资回报，目前仍存在巨大不确定性。标普对甲骨文2027 年后的财务预测缺乏信心，叠加其自身持续疲软的财务表现与尚不明朗的盈利路径，人们难免心生疑问：甲骨文凭什么能维持投资级信用评级？至少我心存疑问。（财富中文网）
凯文·科哈基（Kevin Koharki）拥有工商管理硕士、哲学博士学位，现任普渡大学副教授，也是CAE咨询有限责任公司创始人。二十余年来，凯文一直为包括《财富》100强企业在内的全球大型企业提供咨询服务并发表主题演讲，覆盖银行、保险、分销、制造、航空航天与国防以及法律等行业。作为备受信赖的金融分析师和教育工作者，凯文分析过数百家企业，也曾担任并购分析师。他曾在宾夕法尼亚州立大学、圣路易斯华盛顿大学和普渡大学等顶尖学府为数千名高管、硕士和本科生授课。
Fortune.com上发表的评论文章中表达的观点，仅代表作者本人的观点，不代表《财富》杂志的观点和立场。
如果我告诉你，一家企业未来数年营收预计增长240%、债务飙升410%，几乎无法产生正向现金流，而债权投资者依旧不改对公司的评级，你会相信吗？如果我再告诉你，这家企业正大举投入一项尚未产生足够投资回报的技术，试图实现转型，你会更有信心吗？我也不会。
这就是甲骨文当前的人工智能投资叙事，以及标普全球评级（S&P）给出的评级。
三大评级机构（标普、穆迪、惠誉）为各类组织提供信用评级。投资者据此评估企业的融资成本。各机构采用相近的评级体系，衡量债权投资者无法收回出借资金的概率。其中，低风险企业获得投资级评级（AAA、AA、A、BBB），高风险企业则获得投机级评级（BB、B、CCC、CC、C、D）。
简言之，企业评级越低，融资成本越高。债务投资者只能收回本金及利息，没有额外收益空间。因此他们更关注资本回收率，而非资本回报率。
在研读标普2026年7月9日将甲骨文信用评级下调至BBB-（投资级评级中的最低等级）的公告后，我心中的疑问远多于答案。如前所述，标普预测2022财年至2028财年甲骨文营收将增长239%，债务与现金流亦将增长400%–450%。然而遗憾的是，在持续布局人工智能业务后，其实际现金利润（如自由现金流）预计反而下降32%，且2025至2027年间将持续为负。
标普在2026年7月13日的“甲骨文评级下调解读”电话会议中指出，甲骨文和SpaceX“无疑都是投资级评级中的特例”。标普进一步指出，尽管甲骨文目前的信用指标尚未达到投资级标准，但“维持其投资级评级的原因在于，我们认为随着人工智能业务规模扩大，甲骨文能在合同的第三、第四、第五年实现现金回流。因此，我们仍给予企业时间验证商业模式，在此期间，我们将获得更多数据点，对甲骨文人工智能业务前景更有信心”。
需要说明的是，我并不反对评级机构给予企业时间验证商业模式，但相关数据必须经得起推敲。
出乎意料的是，2022至2028年间，甲骨文债务规模预计增长412%，而利息支出仅增长271%。只有债务利率大幅下行，才会出现这种情况。若甲骨文财务状况持续恶化，这种情况不该出现。信用违约互换（CDS）市场也印证了这一点：5年期CDS合约利差近期超过200个基点，上一次达到该水平还是2008年全球金融危机期间。
倘若标普对甲骨文的预测——尤其是2027年之后的预测——充满信心，那么结论将更具说服力，但事实并非如此。标普坦言：“甲骨文现已进入‘用业绩说话’的阶段，前景不明朗，且仍存诸多疑问。”
这种不确定性部分源于甲骨文的商业模式转型及其与OpenAI的关系。标普将这一新商业模式定性为重资产模式，认为其属于“缺乏护城河的业务”，即不具备核心竞争优势。既然如此，标普为何一边预期甲骨文在2028年前实现营收与会计利润大幅增长，另一边却警示其“盈利路径存在不确定性”？
这种不确定性也源于很难预测甲骨文为实现其人工智能雄心所需投入的资金规模。
具体而言，在与甲骨文沟通后，标普不得不将2027年资本支出指引从600亿美元上调至950亿美元，涨幅近60%。标普坦言对此感到无奈，称其需要“不断追赶”持续攀升的资本开支预测。谁不是呢？
标普当前的信用评级与稳定展望，基于两个前提：一是甲骨文将竭力维持投资级评级，二是公司未来可能通过增发股权稳定资产负债表。标普提示，如果甲骨文无法维持或降低当前债务/息税折旧摊销前利润比率，或是未能在2028年产生正向自由现金流，其评级恐将面临下调压力。颇具讽刺意味的是，从当下到2028年，甲骨文预计只有2028年能实现正向自由现金流。正如我们之前讨论的，要达成这一目标，诸多环节都必须完全符合预期。
人工智能企业能否创造可观投资回报，目前仍存在巨大不确定性。标普对甲骨文2027 年后的财务预测缺乏信心，叠加其自身持续疲软的财务表现与尚不明朗的盈利路径，人们难免心生疑问：甲骨文凭什么能维持投资级信用评级？至少我心存疑问。（财富中文网）
凯文·科哈基（Kevin Koharki）拥有工商管理硕士、哲学博士学位，现任普渡大学副教授，也是CAE咨询有限责任公司创始人。二十余年来，凯文一直为包括《财富》100强企业在内的全球大型企业提供咨询服务并发表主题演讲，覆盖银行、保险、分销、制造、航空航天与国防以及法律等行业。作为备受信赖的金融分析师和教育工作者，凯文分析过数百家企业，也曾担任并购分析师。他曾在宾夕法尼亚州立大学、圣路易斯华盛顿大学和普渡大学等顶尖学府为数千名高管、硕士和本科生授课。
Fortune.com上发表的评论文章中表达的观点，仅代表作者本人的观点，不代表《财富》杂志的观点和立场。
Would you believe me if I told you that debt investors will not change their viewpoint of a company who over the next few years is expected to increase revenue by 240%, debt by 410%, and barely generate a positive cash profit? Would you feel more confident if I told you this company is changing its business by heavily investing in technology that has not yet produced an adequate return on investment? Me neither.
Welcome to the current state of the AI thesis and the ratings decision provided by S&P Global Ratings (S&P) on Oracle.
But first, a little background.
The Big Three rating agencies (e.g., S&P, Moody’s Ratings, and Fitch Ratings) assign credit ratings to organizations. Investors use these ratings to set firms’ borrowing costs. The agencies utilize a similar rating scale, which highlights the likelihood of debt investors not earning back the money they lend to borrowers. Low-risk firms receive an investment-grade rating (e.g., AAA, AA, A, and BBB), while high-risk firms receive a speculative-grade rating (e.g., BB, B, CCC, CC, C, and D).
Put simply, the worse an organization’s rating becomes, the higher the firm’s borrowing costs. After all, debt investors can only earn their principle back plus interest. There is no additional upside. This sharpens their focus on return of capital rather than return on capital.
After examining S&P’s July 9, 2026, decision to downgrade Oracle’s credit rating to BBB- (the last investment-grade rating possible), I had more questions than answers. As I previously mentioned, S&P notes that revenue from fiscal years 2022 thru 2028 should increase 239%. Similarly, both debt and cash flow should increase 400 – 450%. Unfortunately, the actual cash profit Oracle is expected to generate after it invests in AI (e.g., free cash flow) declines 32% and is negative from 2025 through 2027.
S&P noted during its July 13, 2026 “Oracle Downgrade Explained” call that Oracle and SpaceX are “both no doubt outliers for investment-grade”. S&P further stated that while Oracle’s credit metrics are not investment-grade today, “what keeps it investment-grade is that we think that as the AI business scales Oracle should be harvesting cash flow at years three, four, five of their contracts. So we are still giving the company time to prove their business case and over that timeline we will, we will have more data points, more confidence about Orache’s AI business prospects”.
To be clear, I am not opposed to credit rating agencies giving companies time to prove their business models. However, the numbers must also make sense.
Miraculously, from 2022 through 2028, Oracle’s interest costs are forecast to rise 271% while debt increases 412%. This can only occur if interest rates charged on debt decline substantially. This should not happen if Oracle’s financial condition worsens. The credit default swap (CDS) market agrees as spreads on five-year CDS contracts were recently greater than 200 basis points, a level last reached during the 2008 Global Financial Crisis.
It would be equally helpful if S&P was confident in their Oracle forecasts, particularly post-2027, but this is not the case. “Oracle is now a ‘show me’ story with limited visibility and lots of question marks out there,” S&P stated.
Part of this uncertainty stems from Oracle changing its business model, as well as its relationship with OpenAI. S&P describes this new business model a capital-intensive, “no moat business.” In other words, it has no competitive advantage. Why then does S&P expect Oracle to drive considerable revenue and accounting profit growth through 2028, while warning of an “uncertain path to profitability?”
Another contributor is the difficulty in forecasting the investment required for Oracle to achieve its AI ambitions.
Specifically, after discussions with Oracle, S&P had to increase its 2027 capital expenditure guidance almost 60% from $60 billion to $95 billion. S&P notes its frustration by stating it is routinely “playing catch up” regarding ever-increasing capital expenditure forecasts. Who isn’t?
S&P’s current credit rating and stable outlook are predicated on the Oracle’s focus on maintaining an investment-grade rating, coupled with the potential for future equity issuances to stabilize its balance sheet. S&P notes that the rating could be pressured if Oracle fails to maintain or lower its current level of debt-to-EBITDA OR fails to generate positive free cash flow in 2028. Ironically, between now and then, 2028 is the only year Oracle is expected to generate positive free cash flow. As we have already discussed, much must go exactly right for this to occur.
Given the current level of uncertainty regarding the ability of AI companies to generate meaningful ROI, S&P’s limited confidence to forecast Oracle’s financial performance past 2027, and Oracle’s weakening financial performance and uncertain path to profitability, one must wonder how Oracle deserves an investment-grade credit rating. I know I am.
Kevin Koharki, MBA, PhD, is an Associate Professor at Purdue University and founder of CAE Consulting, LLC. For more than two decades, Kevin has advised and delivered keynote speeches for some of the world’s largest firms, including Fortune 100 organizations, in industries such as banking, insurance, distribution, manufacturing, aerospace and defense, and law. A trusted financial analyst and educator, Kevin has analyzed hundreds of companies throughout his career, including during his time as an M&A analyst. He has taught thousands of Executive, Masters, and Undergraduate students at leading universities such as Penn State, Washington University in St. Louis, and Purdue.
The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.